R. A. Farley, University of Liverpool, School of Biological Sciences, Nicholson Building, Liverpool, L69 3BX (fax 0151 7945094; e-mail RAF4@liverpool.ac.uk).
1 Plants can respond to nutrient-rich patches by proliferation of roots or by increased rates of ion uptake; such patches are therefore considered to be an important source of nutrients for plant growth. However, little is known about the spatial and temporal heterogeneity of nutrient concentration at scales that are applicable to the development of root systems of herbaceous perennial plants.
2 A woodland site in North Yorkshire (UK) was studied to measure the scale and extent of nutrient heterogeneity. Phosphate, nitrate, ammonium, pH and soil moisture were measured by destructive soil coring over a period of 2 years, and phosphate, nitrate and ammonium were also measured at a smaller scale using Rhizon Soil Solution Samplers.
3 Soil moisture was relatively constant at 0.6 g g–1 fresh soil weight. Temporal variation was found in all other variables, but the timing of peaks in nutrient concentration was not predictable.
4 Soil coring showed differences in phosphate, ammonium and nitrate concentrations at a scale of over 2 m, which may be too large a scale to affect individual herbaceous perennial plants but might be important for trees. However, the soil solution sampling showed two- to fivefold differences in nitrate and ammonium at scales of 20 cm; the root systems of many herbaceous plants will spread over such distances.
5 Peaks in nutrient concentration that occurred in localized areas lasted no more than 4 weeks. Therefore if these nutrient-rich patches are to be utilized by plants, their roots must respond rapidly.
6 The study showed that localized nutrient-rich areas may form an important source of nutrients for plants in some natural habitats, especially when the general levels of nutrient availability are low. Hence the patchy nature of the soil should be considered in root foraging and population studies.
Plants can exhibit plasticity in root growth in response to nutrient heterogeneity ( Crick & Grime 1987; Campbell & Grime 1989; Robinson 1994). Responses to nutrient enrichment can be highly localized ( Drew & Saker 1975), with additional root growth being concentrated within the nutrient-rich area. These observations have led to the concept of active root foraging, with plants concentrating a greater proportion of their root systems in areas of high resource availability ( Hutchings & de Kroon 1994). Better information about the temporal and spatial scales of nutrient heterogeneity in natural habitats is needed in order to interpret the ecological significance of such responses.
Temporal changes in nutrient concentrations in soil are well recorded, but these will only be important to a root if they occur within its lifetime ( Stark 1991). In non-agricultural soils in temperate ecosystems there is generally a flush of nutrients in the spring and there may also be a smaller increase in nutrient availability in the autumn or winter ( Chapin 1980). Taylor et al. (1982) found seasonal differences in nitrate concentration in both an acidic and a calcareous soil, with a seasonal peak in early spring, which decreased rapidly in summer. In arable and grassland soils, in Denmark, inorganic phosphate decreased in autumn and winter, before starting to rise again in spring ( Magid & Nielsen 1992). Soil pH is also variable in time: in a study of woodlands in northern England, oak and larch litters were shown to have significant monthly variations in pH although no variation was detected in the mineral soil underneath them ( Frankland et al. 1963 ). Therefore, many plants will experience differences in nutrient concentrations throughout the growing season, and may also experience changes in soil pH.
Variation in soil resources also occurs down the soil profile ( Gupta & Rorison 1975) and, depending on the thickness of the soil horizons, distinct zones with significantly different concentrations of nutrients may be millimetres or decimetres apart ( Stark 1991). Merryweather & Fitter (1995) found a rapid decline in phosphate concentration with depth in a woodland soil, ranging from 30 µg ml–1 soil water in the O horizon, at a depth of 1–6 cm, to less than 0.1 µg ml–1 at a depth of 18 cm, in the Bs horizon. As roots grow down through soil horizons, nutrient concentrations typically decline, often to very low values.
Differences in resource concentration also occur spatially across a site. Large-scale differences have been recorded in both uncultivated soils (at a scale of 7–26 m) and cultivated soils (at a scale of 48–108 m; Robertson et al. 1993 ); in cultivated soils, ploughing mixes the soil and decreases heterogeneity. However, environmental heterogeneity will only affect plant growth if it occurs at scales that are relevant to, and detectable by, plants. The smallest scale of significant nutrient heterogeneity measured by Robertson et al. (1993) was 7 m; while variation at this scale may be experienced by the roots of trees and shrubs it will be of little importance to smaller herbaceous plants. If heterogeneity occurs at very small scales or for only very short periods of time, then no morphological response would be expected ( Robinson 1996).
Variation in the availability of soil resources has also been measured over spatial scales at or below the scale of individual plants. Snaydon (1962) found significant variation in soil calcium, phosphate and potassium concentration over distances of 60 cm in acid low-phosphate soil. Jackson & Caldwell (1993a) found that nitrate and ammonium concentrations varied by two to three orders of magnitude within a 12 × 10-m plot, and that as much variability in ammonium, nitrate, phosphate and potassium concentrations occurred in a 1-m2 plot as in the whole 120-m2 site. Gross et al. (1995) sampled nitrate within 12 × 20-m plots in an abandoned field and found significant variation in nitrate concentration in the surface 5 cm of the soil over a distance of 0.55 m.
Woodlands are often nutrient-poor habitats; large trees are major sinks for all nutrients and their mycorrhizal networks may effectively close the mineral nutrient cycle ( Grime 1991). Gross et al. (1995) found that soil nitrogen concentration was six times lower in a woodland site than a newly abandoned field. Organic horizons have high carbon : nutrient ratios ( Arp 1984), leading to nutrient deficiency. McNabb et al. (1986) found considerable variation in surface nitrogen and carbon in forests in Oregon, USA, although they did not measure the scale of this variation. Fine root growth, which is usually associated with nutrient uptake, occurs predominantly in the upper 20 cm of forest soil ( Hendrick & Pregitzer 1996). Therefore in this type of low-nutrient environment, nutrient-rich patches and pulses in the upper soil horizons may be an important source of nutrients for plant growth, and it is important to establish the scales of nutrient heterogeneity experienced by woodland plants.
The objectives of this study were to answer the following questions:
1 What is the degree and extent of spatial and temporal variation in soil nutrient concentrations at a single woodland site?
2 Are there seasonal patterns in nutrient variation, i.e. are peaks in nutrient availability predictable each year, and do they occur at times of year when they can be utilized by growing plants?
3 What is the spatial scale of nutrient variation? Does it occur at scales that are likely to be relevant to the ground flora at the site?
Once the pattern and predictability of nutrient variation at a particular site has been established, the root growth of herbaceous plant species at the site could be studied at similar spatial and temporal scales to investigate the importance of soil heterogeneity in nutrient foraging ( Farley & Fitter 1999).
The study site
The study site was Pretty Wood, Castle Howard, North Yorkshire, UK (grid reference SE 732 687, altitude 50 m a.s.l.). The soil is a well-drained sandy brown earth with a rich humic horizon above a highly leached, nutrient-deficient Bs horizon ( Merryweather & Fitter 1995). The area was planted 250–300 years ago and the tree canopy is oak, mainly Quercus petraea (Mattuschka) Liebl., with some Q. cerris L. and Acer pseudoplatanus L. ( Merryweather & Fitter 1995).
Four 1-m2 quadrats were marked out at 2-m intervals along a line transect down a slight slope in the spring of 1993, to monitor large-scale spatial variation and seasonal patterns in nutrient concentration. Each quadrat was subdivided into four equal subquadrats, which were further subdivided into 25 squares, each 10 × 10 cm. The quadrats were all under the tree canopy, and had little ground vegetation throughout the study. To investigate seasonal patterns in nutrient concentration, samples were taken from different locations within the quadrats over a period of 2 years. Beginning in April 1993 there was a monthly destructive harvest, at which a 6-cm diameter core was removed from one randomly selected 10 × 10-cm square from each subquadrat, to give four replicate samples per quadrat. The sample hole was filled with soil taken from nearby to avoid leaving holes that might affect water drainage and soil temperature. The L and F horizons (approximately the top 2–3 cm) were discarded, and the H (approximately 1–2 cm deep) and A (approximately 3–4 cm deep) horizons were placed in separate plastic bags to prevent moisture loss before laboratory analysis.
One gram of fresh soil was dried at 80 °C for 48 h, and the water content of the wet soil was then calculated. Further samples were mixed with water (4 g of fresh soil to 20 ml deionized water; Allen 1974, p. 42) then mixed end-over-end for 30 min before 0.5 ml of 0.1 m boric acid was added to stop microbiological activity, and the samples were filtered through Whatman no. 42 ashless filter paper (Whatman International, Maidstone, Kent) overnight in a cool room (≤ 17 °C). Ammonium (last 14 months only) and nitrate (all dates) were measured using Russell ion selective electrodes (ISE/93-3139 and ISE/93-3079, respectively) with a Russell reference electrode (900029) (Russell pH Ltd, Cupar, Fife) attached to a Corning volt meter (Corning Costar (UK) Ltd, High Wycombe, Bucks). Phosphate was measured colorimetrically by the ascorbic acid–molybdenum blue method ( Allen 1974, p. 135) at 882 nm on a Hitachi U-1100 spectrophotometer (Hitachi Scientific Instruments, Ratingen, Germany). Concentrations were expressed as mmol l–1 soil water. Another sample of soil was mixed with a neutral 0.01 m CaCl2 buffer to form a slurry ( Allen 1974, pp. 17), and soil pH measured to two significant figures with a Russell pH electrode (CE7 L).
Spatial variation in soil pH, soil water content, ammonium, nitrate and phosphate concentrations and N:P ratio was analysed in SPSS (version 6.1) using nested multivariate analysis of variance, so that quadrat was tested against subquadrat, with month (April–March) of data as the dependent variables. As there were full data sets (24 months) for nitrate and phosphate, seasonal effects could be looked for by including year of sampling as a factor for these variables. Effects of quadrat, soil horizon and the interactions between these terms were tested for each soil variable. Significant differences detected by multivariate analysis of variance (Pillai’s trace test; Eaton 1983) could be due to differences in just 1 month. Therefore univariate tests were used to indicate the months in which significant differences were occurring, and specific quadrat differences were identified using Bonferroni’s 95% confidence intervals ( Morrison 1976). To test for seasonal patterns in nutrient variation, the means for the 2 years were compared by Spearman’s coefficient of rank correlation ( Steel & Torrie 1980): monthly values for each year were ranked and differences between ranks from the paired variables compared.
Rhizon soil solution samplers
The permanent quadrats were 2 m apart, and thus provided information on variation at a spatial scale larger than that at which roots of herbaceous plants would be likely to forage. Even subquadrats only tested for differences in soil properties at a half-metre scale. Therefore, to measure whether variation in nutrient concentrations occurred at the scale of individual plants, a further sampling area was set up in March 1994. Rhizon Soil Solution Samplers (Rhizon SSS; Eijkelkamp Agrisearch Equipment, Giesbeek, The Netherlands) were used to measure small-scale change, because they allowed repeated sampling of the same area of soil and so gave an indication of nutrient patch longevity. Rhizon SSS are miniature porous cups with a 2.5-mm diameter and 10-cm long hydrophilic porous polymer tube, connected to the soil surface by a PVC tube. Soil water can be extracted from the soil surrounding the Rhizon SSS by creating a vacuum at the end of the PVC tube.
The Rhizon SSS needs to be inserted horizontally into the soil in order to give good soil contact and improve sample collection, so the quadrats were set up in the side of a steep bank, about 50 m from the permanent quadrats. Three 20 × 20-cm quadrats were marked out in a horizontal line, 20 cm apart. Each quadrat was divided to give four subquadrats, 10 × 10-cm square, with three Rhizon SSS per subquadrat. To avoid damaging the Rhizon SSS on roots or stones, a hole 10 cm deep was first made with a 2-mm diameter wire, and the Rhizon SSS then inserted. To extract samples, a syringe needle was fixed to the end of the Rhizon SSS sampler tube. This was pushed through a resealable (suba-seal) bung into a vacuumed flask. The vacuumed flasks contained 0.5 ml of 0.1 m boric acid to stop microbial activity. The flasks were collected every 2 weeks. Any liquid in the flasks was analysed for phosphate, nitrate and ammonium concentration as before. Where necessary the soil water samples were diluted to give sufficient volume for analysis.
Results were expressed as mmol l–1 and analysed by nested multivariate analysis as before in SPSS, with sampling times as the dependent variables. Effects at both quadrat and subquadrat scales could be tested, because the three samples within each 10 × 10-cm subquadrat were replicates. At many of the sampling times there were missing data, as not all flasks contained sufficient soil water for analysis and some contained no sample as the vacuum had failed. Therefore, only a subsample of the data was analysed.
At the last sampling time for the Rhizon SSS, 2-cm wide soil cores of the soil surrounding the samplers were taken, to the same depth as the samplers. These cores were then processed in the same way as the cores from the permanent quadrats. Results obtained from the two methods were compared by analysis of variance in Minitab (Release 9), using all data as replicates.
Mean soil water content over the 2-year period was 0.57 g g–1, and differed significantly from this value only in September of both years, when it fell to 0.47 and 0.51 g g–1, respectively, and in May 1993, when it was 0.46 g g–1. Although multivariate analysis of variance showed a significant difference in water concentration between years (F12,27 = 3.182, P = 0.006), this was largely due to the lower values of soil water content in May and September 1993.
Soil pH also differed between years (F12,37 = 57.26, P < 0.001); there was a significant difference (P < 0.05) in pH between years in all months except January (F1,48 = 3.67, P = 0.061). However, the range of measured soil pH was narrow (2.9–3.4), with 50% of all measured values being either pH 3.1 or 3.2. Within-site variation for pH was very small.
The concentration of soil phosphate was lower in the second year than in the first (F12,32 = 11.51, P < 0.001; Fig. 1a). In April–May in the first sampling year there was a peak in phosphate concentration that did not occur in the second year. In both years, the lowest concentration of phosphate occurred in October. When values for the 2 years were compared there was no consistent seasonal pattern in phosphate concentration (r10 = 0.53, P < 0.1). However, if only the values from September–March were considered, a strong seasonal pattern could be detected (r5 = 0.89, P < 0.01).
Soil nitrate concentration also showed a large difference between years (F12,33 = 35.47, P < 0.001; Fig. 1b), with significant differences between years in all months except January and February. There was no consistent seasonal pattern of nitrate concentration when values for the 2 years were compared (r10 = 0.55, P < 0.1). As there were not 2 full years of data for ammonium, yearly differences and seasonal patterns could not be analysed. The temporal patterns of nutrient concentration were similar for nitrate and ammonium in 1994–95 ( Fig. 1c), although peaks in ammonium concentration occurred a month later than those for nitrate. There was a strong correlation between mean nitrate concentration and mean ammonium concentration 1 month later (r10 = 0.92, P < 0.001).
The total nitrogen (ammonium plus nitrate) to phosphate (N:P) ratio for 1994–95 ( Fig. 1d) varied throughout the year. The lowest N:P ratios occurred in May and again from September to November. The N:P ratio was higher from June until August.
Spatial variation in soil conditions and soil nutrients also occurred in both the horizontal and vertical dimensions. On a vertical scale there were significant differences in soil water content between the soil horizons (F12,32 = 40.66, P < 0.001), with the H horizon being significantly wetter than the A horizon in all months except June. Phosphate and nitrate concentrations also varied between horizons (F12,32 = 17.10, P < 0.001; F12,33 = 5.86, P < 0.001, respectively), but this variation followed no obvious seasonal pattern.
Spatial variation in soil nutrient concentrations occurred across the site, with significant differences between quadrats for all measured nutrients. Although ammonium concentration showed significant spatial variation (F36,42 = 1.86, P < 0.025), Bonferroni comparisons indicated that this variation was only significant in some months, with one or more quadrats deviating significantly from the monthly mean ( Fig. 2c). In contrast, there was variation between quadrats in most months for phosphate (F36,102 = 5.28, P < 0.001; Fig. 2a) and nitrate (F36,102 = 5.28, P < 0.001; Fig. 2b) concentrations, with the phosphate concentration generally being significantly lower than the mean site concentration in quadrat 3 in most months. In August and March, phosphate concentrations were significantly different from the monthly mean in all quadrats. For nitrate, there were differences between quadrats in eight out of 12 months. Again values for quadrat 3 were consistently lower than for the other quadrats. The N:P ratio differed significantly between quadrats (F36,102 = 2.09, P < 0.001; Fig. 2d), with the lowest N:P ratio generally being found in quadrat 4 and the highest N:P ratios in quadrat 3. There was considerable within-quadrat variation for all nutrient concentrations and for the N:P ratio, particularly in July and August. The differences in nutrient concentration between quadrats demonstrated spatial variation at a scale of 2 m.
Small-scale spatial variation
At the smaller scale investigated using Rhizon SSS, there was variation in both ammonium (F14,18 = 2.00, P = 0.007) and nitrate (F14,18 = 2.66, P = 0.026) concentrations ( Fig. 3), with significant differences in ion concentration occurring between quadrats that were only 20 cm apart. There were two- to 10-fold differences in nutrient concentration between adjacent quadrats for nitrate, and three- to fivefold differences for ammonium. No significant variation occurred at the quadrat level for phosphate, or at the subquadrat level for nitrate, ammonium or phosphate. Significant variation could therefore be detected at a scale of 20 cm for some ions but not at a scale of 10 cm for any ion.
Small-scale temporal variation
There were large differences in soil nutrient concentration in the Rhizon SSS between sampling times ( Fig. 3), but peaks in nutrient concentration were generally short-lived. Nitrate showed the highest concentration in quadrat 3; when nitrate was at its peak concentration in quadrat 3 it was 10 times greater than the concentration recorded in quadrat 2. The peak in nitrate concentration lasted a month, with the highest values being recorded over 2 weeks. The nitrate concentration dropped rapidly after mid-September 1994, which suggested that changes in nutrient concentration may be very rapid. There was no relationship, either direct or lagged (1–3-month lags were tested), between nitrate or ammonium concentrations in the samples at this smaller spatial scale.
The Rhizon SSS results are not directly comparable to those obtained from soil coring. When water extracts, collected from soil cores adjacent to the Rhizon SSS at the same time as the final samples were taken, were compared with values from Rhizon SSS, they had significantly higher concentrations of nutrients ( Fig. 4). Phosphate concentrations were approximately 18 times greater in the soil core samples than in the Rhizon SSS, ammonium concentrations seven times greater and nitrate concentrations three times greater.
There was considerable temporal variation in soil nutrient concentration in Pretty Wood. Nutrient concentrations varied throughout the year, with the general trend in seasonal variation being similar to those recorded by soil coring in other woodland studies ( Frankland et al. 1963 ; Davy & Taylor 1974); nutrient concentrations generally increased during the spring or summer, fell in the late summer, then showed a slight increase in the autumn. There was a main peak of nitrate concentration in Pretty Wood in July in both years, a month later than measured by Davy & Taylor (1974), whereas phosphate peaked in April and May, in the first year, a month earlier than the phosphate peak recorded by Frankland et al. (1963). Leaf litter represents a major nutrient input for woodland soil, with the nutrient concentration of the soil being affected by the state of decomposition of the litter. Net mineralization is usually greater during the spring and summer than in the winter ( Ehrenfeld et al. 1997 ) because the soil is warmer, favouring microbial activity. Leaching from the fresh leaf litter ( Chapin 1980) may have caused the rise in nutrient concentration in the autumn. The differences in the timing of the spring and summer peaks in nutrient concentration between sites could be due to differences in litter quality, temperature, soil pH and mineralization rates between sites. However, although nutrient concentrations appeared to follow a seasonal trend, there was no strong evidence that these seasonal patterns were consistent in consecutive years, possibly due to variation in temperature and the quality and quantity of the litter fall.
The observed range in nutrient concentrations was high: monthly mean nitrate and ammonium concentrations, measured by soil coring, had a sevenfold and fourfold range in the second year of sampling, respectively, and the range for nitrate was even greater in the Rhizon SSS. Similar seasonal flushes in nutrient availability have been seen in other studies of soil nutrient status, both in forests ( Garten et al. 1994 ) and in grassland ( Gupta & Rorison 1975), suggesting that the root systems of most plants will experience different nutrient availabilities over the year. Some soil samples from Pretty Wood had very low nutrient concentrations, suggesting that nutrients are likely to limit plant growth throughout the year, and that localized nutrient-rich patches may form an important source of nutrients for the woodland plants. In addition, nutrient-rich patches varied in quality, with areas rich in nitrogen often being low in phosphorus. The N:P ratio of the patch might affect whether it is exploited by plant roots ( Cui & Caldwell 1997).
The soil pH at Pretty Wood was 3.1, as previously reported by Merryweather & Fitter (1995), and this is low compared with other British oak woodlands (pH 4–4.5, Frankland et al. 1963 ; pH 4.5, Brown 1974) and to forests in north central USA ( Grigal et al. 1991 ), possibly partly due to differences in methods for recording. At the very low pH recorded in Pretty Wood, mineralization would have been restricted, and could have limited soil nutrient concentrations during the year. The soil water content in Pretty Wood showed little variation, and was similar to that of other British woodlands ( Frankland et al. 1963 ).
Nutrient concentrations varied across the sampling area, and all nutrients showed some differences in concentrations between quadrats. Soil coring showed that there was significant spatial variation in nutrient concentration at a scale of 2 m. This is similar to the scale at which significant variation was found by Lechowicz & Bell (1991) in a Canadian forest, but greater than the 1-m scale at which soil nutrients were found to vary significantly in a loam soil ( Jackson & Caldwell 1993b). Variation in nutrient availability on this scale may affect the distribution of species within a community ( Ehrenfeld et al. 1997 ) if species differ in their tolerance of resource limitation. Although such large-scale variation is unlikely to be relevant to herbaceous plants, it may be important for trees and shrubs, the roots of which spread over several metres, and which tend to have higher root densities in the surface soil ( Roberts 1976; Hendrick & Pregitzer 1996).
However, we also detected variation in nutrient concentration at a much smaller scale, 20 cm, which is more likely to be relevant to herbaceous plants, although not at a still smaller scale, 10 cm. A three- to fivefold difference in ammonium and a two- to 10-fold difference in nitrate concentration were measured between quadrats by the Rhizon SSS. Similar (twofold) differences in nitrogen concentration have been shown to increase ion uptake after 3 days from nutrient-rich patches compared with the surrounding soil ( Jackson & Caldwell 1991). Such an ability to respond quickly to nutrient-rich areas is likely to be important for the species found in Pretty Wood, as localized peaks in nutrient concentration here were only 2–4 weeks in duration. Nutrient heterogeneity in Pretty Wood appears to occur at spatial and temporal scales that some plants may be capable of exploiting, either by increased nutrient uptake ( Robinson & Rorison 1983; Jackson & Caldwell 1991) or by localized root proliferation ( Eissenstat & Caldwell 1988; Robinson 1994).
The two methods of collecting nutrient samples (soil coring and Rhizon SSS) were not directly comparable, as the Rhizon SSS gave lower concentrations than water extracts from soil cores taken from the same area. Given that the Rhizon SSS sampled soil water over a period of time from an unknown volume of soil, whereas the extraction from the soil cores indicates the conditions within the soil on a specific day, this difference is to be expected. However, the patterns of variability were the same for both sampling methods. The concentrations obtained by the two sampling methods were most different for phosphate, the immobility of which means that sampling even by extraction of soil water probably greatly overestimates the amount of phosphate available for uptake by roots. The Rhizon SSS measurements probably gave a better indication of nutrient concentration available for plants, as plants obtain nutrients from the soil solution.
The results indicated that any plant which produces roots with a horizontal spread of 20 cm or more will experience both temporal and spatial variation in soil nutrients. Woodland plants such as Glechoma hederacea, Veronica montana and Silene dioica are common at the site, and these plants can produce over 10 m of root in 6 weeks ( Farley 1996). At typical root length densities for this habitat (approximately 1 cm cm–3), this length of root could occupy at least an area with a radius of 5–10 cm to a depth of 10 cm; therefore different parts of a single root system will often be exposed to different nutrient concentrations. Many of the peaks in nutrient concentration occurred at times of active plant growth, but variation occurred throughout the year. Localized nutrient-rich areas may be an important source of nutrients for plant growth ( Chapin 1980), particularly at the times of year when the general levels of nutrient concentration in the woodland soil are low.
Numerous studies ( Drew & Saker 1975; Fitter 1982; Campbell & Grime 1989) have shown that plants can exhibit morphological plasticity in root growth in response to localized nutrient heterogeneity. This plasticity acts to increase nutrient uptake in localized nutrient-rich areas, usually by an increase in root growth. As this study shows that small-scale nutrient heterogeneity does occur in Pretty Wood at times and scales applicable to these studies, an ability to respond to localized nutrient-rich areas would appear to be an advantageous trait for woodland plants. However, the results also indicate that there is likely to be a great diversity in the types of nutrient-rich areas that plant roots may encounter. Nutrient-rich areas varied in their duration, N:P ratio and in the ion that was in highest concentration. While some of these areas had high concentrations of nutrients in them, others were only slightly higher than the surrounding soil. The results indicated that nutrient-rich patches were relatively short-lived, so that a plant must be able to respond quickly if it is to exploit the nutrients available in them. The benefit that a plant may get from exploiting a patch will depend not only on the concentration of nutrients in the patch, but on the nutrient status of the plant, and the levels of ion availability in the surrounding soil. At times of the year when nutrient concentrations are very low, localized nutrient-rich areas may be an important source of nutrients for plant growth. Predictions about the growth of species, particularly those adapted to low-nutrient habitats, should therefore take account of the patchiness of the environment in which the species live.
We are grateful to Castle Howard Estate for access to the site, to James Merryweather for field assistance, and to M. Hutchings and B. Campbell for their help with earlier drafts of this manuscript. This project was funded by the Natural Environment Research Council.
Received 29 October 1998revision accepted 4 March 1999