of small-scale soil disturbances in the Negev Desert of Israel. Received: 16 August 1996 / Accepted: 26 October 1997. Abstract We investigated whether plant ...
Oecologia (1998) 114:106±117
Ó Springer-Verlag 1998
Bertrand Boeken á Clive Lipchin á Yitzchak Gutterman Noel van Rooyen
Annual plant community responses to density of small-scale soil disturbances in the Negev Desert of Israel
Received: 16 August 1996 / Accepted: 26 October 1997
B. Boeken (&) á C. Lipchin á Y. Gutterman Jacob Blaustein Institute for Desert Research and Dept. of Life Sciences, Ben-Gurion University of the Negev, Sede Boker Campus 84990, Israel
decreased with increasing digging density, but only on one of the slopes. At the highest digging density, plant density and species number in the diggings did not decrease down the slope, as expected if interference between diggings in runo water capture were the cause of the digging density eect. There was a weak decrease in biomass production in 1994±1995 down the slope. We used a simple mathematical model to estimate whether the distribution of rainfall intensities that occurred during the winter of 1994±1995 could result in dierences between digging densities in the amount of water captured by the diggings, and whether this could explain the observed eect of digging density. The model showed that there were four events during which less water was captured by the diggings at high digging densities, except in the topmost row of diggings. Soil moisture measurements, however, showed very little dierence between diggings at dierent digging densities. We explain our ®ndings as the result of the interaction between the properties of the disturbance patch with its surroundings, as the diggings capture resources in the form of runo water, and seeds moved primarily by wind. The additional resources and seeds captured in diggings increase plant density, species richness and productivity relative to the undisturbed matrix. However, the contrast in plant responses between the disturbed patches and undisturbed soil diminishes at higher digging densities. We explain this as interference among diggings at close proximity. As we did not detect a decrease in plant responses down the slopes, we conclude that interference is due to interception of the winddriven, non-directional ¯ow of seeds. Interception of the down-slope ¯ow of runo water by upslope diggings is insucient to aect plant density, determined at the beginning of the season. Later in the season, runo interception may become important for biomass production.
N. van Rooyen Department of Botany, University of Pretoria, Pretoria, 0002, South Africa
Key words Plant productivity á Plant density á Slope direction á Soil disturbance á Species richness
Abstract We investigated whether plant diversity and productivity in small-scale soil disturbances, which is known to be higher than in undisturbed soil, decreases as the density of the disturbances increases. We studied this in an experiment with soil diggings (15 cm diameter and 15 cm depth) dug at a range of densities, on a north- and a south-facing slope of a watershed in the central Negev Desert of Israel. The diggings were similar to the commonly occurring pits made by porcupines (Hystrix indica) as they forage for below-ground plant parts. We used four levels of digging density, within the naturally occurring range in the region, represented by a rectangular plot with rows of diggings dug at four distances between diggings. The plots were laid out in a blocked design with three replications on both slopes, with each block containing all four levels of digging density. In the spring of 1992, 1994 and 1995 we measured plant density, species richness and plant productivity in the diggings, and in adjacent equal-sized undisturbed control areas (``soil matrix'') and on the mounds made by the removed excess soil. Plant density, species richness and productivity of annual plants were higher in the diggings than in the undisturbed matrix, while these responses were very low on the mounds. Plant density, species richness and productivity in the diggings, but not in the matrix or mounds, decreased as digging density increased. This eect varied slightly with location within a watershed and with annual rainfall. The density of seeds captured in the diggings from outside the digging during the 1995 dispersal season
107
Introduction The importance of disturbance for diversity and productivity of biological communities has long been recognized (Grubb 1977; Tilman 1982; Pickett and White 1985). Disturbance creates new opportunities for coexistence of species, due to changes in conditions in situ and in the interactions between the disturbed patch and its surroundings (Pickett and White 1985). Whatever its cause, i.e. by animals, humans or plants (e.g. treefall) (Platt 1975; Gutterman 1988; Huntly and Inouye 1988; Boeken and Shachak 1994; Watt 1947; Brokaw 1985; Canham 1988), or abiotic factors such as storms, ®res and wave action (Spurr 1956; Heinselman 1973; Connell 1978; Sousa 1984), disturbance creates patches in which new individuals can become established (Connell 1978; Pickett and White 1985) and where availability of resources is altered (water, nutrients and/or light: Tilman 1982; Vitousek and Denslow 1986; Canham 1988). Disturbance aects plant communities in two ways: (1) by local changes within the disturbance patch, including the release of locally existing resources (Vitousek and Denslow 1986), and recruitment of plants from propagules already present in the soil (Egler 1954; Grubb 1977; Goldberg 1987); and (2) by eects originating outside the patch, mainly by the arrival of resources (Canham 1988; Gutterman et al. 1990; Shachak et al. 1991; Boeken and Shachak 1994; Boeken et al. 1995) and plant propagules from outside the patch (Reichman 1984; Gutterman 1993; Boeken and Shachak 1994; Boeken et al. 1995). Both local and external processes determine how communities within disturbance patches respond to the presence of other patches (including other disturbance patches) in the vicinity. Here we test the hypothesis that when external eects are important, patches interfere with each other and plant communities are negatively aected by the presence of other ``competing'' patches. We asked the question: what is the eect of the density of disturbed patches embedded in a matrix of undisturbed area (Pickett and White 1985) on density, diversity and productivity of plant communities within disturbance patches? Our hypotheses were that if only local vegetation processes occur in the disturbance patches, then diversity and plant density should be independent of the presence of other disturbed patches. However, if these processes depend on the supply of plant propagules and/ or resources from outside, the responses of the plant communities within the disturbance patches may depend on the presence, extent, and abundance of similar patches in the vicinity. In this case, patches interfere with each other, as interception of a ¯ow by one patch will decrease its capture by other patches. The net result will be that the positive eects of disturbance on the plant communities within the disturbance patches should diminish as the density of these patches increases. We tested these hypotheses in the central Negev Desert of Israel, using a ®eld experiment with small-scale soil disturbances (pits with a diameter and depth of
15 cm) constructed at a range of densities. We used this kind of disturbance patch for several reasons. First, pits with similar dimensions are very common in the Negev. They are made by the Indian crested porcupine (Hystrix indica), as it digs for below-ground parts of geophytes and hemicryptophytes (Gutterman and Herr 1981). Second, the occurrence of the diggings and their eect on the vegetation have been studied for over a decade (Gutterman and Herr 1981; Gutterman 1982, 1987, 1988; Alkon and Olsvig-Whittaker 1989; Gutterman et al. 1990; Shachak and Brand 1991; Shachak et al. 1991; Boeken et al. 1995). Third, their small size facilitates experimental manipulation of properties and densities. Fourth, the fact that the plant communities in the diggings consist mainly of annuals (Gutterman and Herr 1981) makes them suitable for study. Finally, each growing season is a new generation, and the disturbance activity removes the soil seed bank, which is generally super®cial (Gutterman 1994). Therefore, porcupine diggings and their arti®cial equivalents form a convenient model in which to study the eects of disturbance density on plant communities. A number of studies have shown that plant density, productivity and species richness are higher in porcupine diggings than in the undisturbed matrix (Gutterman and Herr 1981; Gutterman et al. 1990; Shachak et al. 1991; Boeken et al. 1995). In this study we focus on the eects of experimental variation of the density of arti®cial diggings on plant density, productivity and species richness of annual plant communities within diggings. We compare these plant responses within diggings to those in the undisturbed matrix, and in soil mounds created by the disturbance. We conducted the experiment on two dierent slopes, and measured the plant community responses in disturbed and undisturbed patches for 3 years between 1992 and 1995. Previous studies (Gutterman and Herr 1981; Gutterman 1982, 1987, 1988; Alkon and Olsvig-Whittaker 1989; Gutterman et al. 1990; Shachak and Brand 1991; Shachak et al. 1991) have attributed the responses of the plant communities in porcupine diggings relative to the undisturbed matrix to greater availability of resources (water, dissolved nutrients, organic matter, soil) resulting from the capture of runo ¯ow. However, recent studies have demonstrated that these responses are also due to greater seed capture in the diggings (Boeken and Shachak 1994; Boeken et al. 1995). Soil pits have been shown to be especially eective as wind traps for seeds (Reichman 1984). In porcupine diggings studied by Boeken et al. (1995), this was illustrated by higher densities of wind-dispersed species in particular.
Materials and methods Site description The study was done in the central Negev, Israel, near Sede Boker (32°52¢N, 34°47¢E) where the climate is strongly seasonal with
108 warm summers and mild winters. Rainfall, which only occurs in winter, averages 100 mm annually (Evenari et al. 1982). Maximum temperatures may be higher than 40°C in July and August. In January and February, minimum temperatures are rarely below freezing. The experiment was conducted on parts of a north and a southfacing slope of the same watershed, 20±30 m above the wadi bed. This part of the slope has a stony colluvial soil with a thickness of up to 2.5 m. Higher on the slopes there are large outcrops of bedrock, which are important for the formation of runo. This occurs during the four to ten rain events per winter which are greater than 8 mm (Yair and Shachak 1987). The vegetation of the slopes consist of scattered shrubs with an average cover of 30±50% (Olsvig-Whittaker et al. 1983), with in winter a large variety of annuals and seasonal perennials (Evenari et al. 1982; Gutterman 1993). Experimental design Soil pits similar to porcupine diggings (15 cm diameter by 15 cm depth) were dug in the winter of 1991±1992. The excavated soil from each pit was deposited down-slope from it forming a small mound, as is also found together with natural digs (Gutterman 1982). The arti®cial diggings were dug at four densities, 0.7, 1.2, 2.8 and 11.1 diggings m)2, with distances between the diggings of 120, 90, 60 and 30 cm respectively. The densities re¯ect the commonly found range of naturally occuring diggings (Alkon and OlsvigWhittaker 1989; M. Shachak, personal communication), although the high densities only occur in areas of 1±2 m2. For each density, diggings were arranged in a rectangular plot of six rows along the slope with ®ve diggings each, with equal distances between them. Each plot contained 30 diggings, with distances between them varying according to density. Along each slope, 12 plots were constructed, in three groups of all four densities, with a distance of 0.5 m between the plots. The groups were used as blocks in a randomized block design with three replications of the four densities, with patches (12 diggings, matrix patches and mounds) nested within each density plot. In each density plot, only the inner 12 ``target'' diggings were used for measurements of plant community responses, to avoid edge eects. The topmost rows of all density plots of all blocks were in line, but because of the dierent distances between rows of diggings, the high-density plots extended less far down the slope than the low densities. The area of all three blocks together on each slope was 42 ´ 8 m. Estimated runo capture in diggings We used a simple geometric model to explore the variation in the amounts of water received by diggings, based on the pattern of rainfall events in 1994±1995. The aim of these calculations is to show dierences in runo water capture between diggings at the experimental densities used, and between the top rows of diggings and the other rows further down the slope. If the model shows great dierences in water capture per digging for dierent digging densities, and if patterns of plant community responses among digging densities and among diggings at dierent positions on the slope coincide with the ®ndings of the model, we can draw conclusions about the importance of runo capture in causing interference among diggings for water at the dierent digging densities. The model required a number of assumptions about the area contributing runo, the portion of rainfall ¯owing down as runo, the minimum rainfall to produce runo, and the volume of the diggings. We assume that diggings receive runo water from an area directly above and to the side, extending sideways to the middle of the distance to the next digging. Thus, diggings below the top row receive runo water from a small area extending upslope to the digging directly above it, in addition to direct rainfall and the runo not captured by the upslope digging. Diggings in the top row
receive runo from a larger area, extending further upslope. We can make only arbitrary assumptions about its length. We use two values: one very small, and one very large. The latter is more realistic, because the extensive rock outcrops a few meters above the plots generate large amounts of runo. The calculations enable us to examine the eects of runo input from above the plot on the likelihood of interference for water among diggings. As rainfall events larger than 8 mm generally produce 20±30% runo (Yair and Shachak 1987), we did our calculations for both values. We further assumed that diggings have an inverted cone shape with depth and diameter of 15 cm, giving a volume of 795 ml.
Soil moisture measurements During the last study year (1994±1995) we took soil moisture measurements in diggings and adjacent matrices of a depth of 0±5 cm with locally installed 3-mm-diameter stainless steel probes using time domain re¯ectometry (Dasberg and Dalton 1985). Unfortunately, only the data for two dates could be used (10 and 15 November 1994), taken in three diggings and matrices in four digging density plots of only one of the blocks. The ®rst measurement was done 4 days after a 3.1-mm rain event, with a cumulative amount of 43.05 mm; during the second there was a 3.1-mm rainfall event, 46.15 cumulative.
Soil deposition in the diggings Over time, loose soil from the mounds and organic matter accumulated in the diggings so that they gradually silted-up. The siltingup rate of the diggings was assessed annually by measuring the length of a wire peg that had been covered since its placement in each digging at the beginning of the experiment.
Plant measurements All individuals of annual plant species were identi®ed and counted in three sample areas at the site of each digging. The sample areas were of three patch types: the digging itself, the excavated soil mound at the base of the digging, and the control area of undisturbed soil matrix adjacent to the digging. Counts were made during the spring of 1992, 1994 and 1995. In 1993 no counts were taken. The vegetation responses determined in this study were total plant density per patch (diggings, mounds and matrix), species number per patch, and species richness, calculated as species number divided by the log of density (Peet 1974). Annual biomass production per digging was also measured, in all outer ``guard'' diggings of the top and bottom rows and in three random inner ``target'' diggings of all digging density plots in spring 1995. All plants present in the diggings were harvested, identi®ed, counted, oven-dried at 75°C and weighed.
Seed densities At the end of the summer of 1995, after the last census in the experiment, we collected the seeds that had accumulated in one of the diggings per density plot from which the biomass samples were harvested. Removal of the vegetation before seed dispersal ensured that only seeds from the outside entered the diggings. The seeds were collected, together with the loose soil and organic material lying on top of the compacted soil within the diggings from before the end of the rainy period. Complete samples were sown on top of sterilized sand in 15-cm-high ¯ower pots, placed under a 50% shade net and mist-irrigated every day. All emerged seedlings were counted and identi®ed by species, giving estimated total density and species number of seeds that are ready to germinate.
109 Statistical procedures Initially we used a repeated-measures ANOVA with random blocks to assess the overall eects of years as the repeated-measures variable, location (north- vs. south-facing slope), digging density, patch type (diggings or matrix, for plant density; also mounds), and their two-way interactions on plant density and species number per patch (Sokal and Rohlf 1995). Individual patches were nested within density plots. Patch type and year eects were tested against individual patches per year, while digging density and location were tested against the block ´ digging density interaction. We ascertained whether species number per patch responded only to plant density, as an artifact of denser samples with higher probability of including more species (Peet 1974), or also independently of plant density. The latter would mean that digging density had a qualitative eect on site availability and quality. We tested species number against plant density per patch with the latter as a covariate. Biomass production was also tested against digging density, plant density and their interaction. The 1995 seed density estimates from the germination trials were analysed in a two-way ANOVA with location (north and south-facing slopes) and digging density. In addition, we attempted to identify the vector of the ¯ow involved in possible interference among diggings (wind or runo) by comparing diggings located higher and lower on the slope. Directional dierences in annual plant density in the diggings per slope were tested with position within the density plot and their interaction as additional variables in the blocked ANOVA model. For the 1995 biomass production data we compared the eect of digging density for the top and bottom rows of guard diggings against the residual mean square. The eect of silting-up of the diggings on annual plant density in the diggings was statistically tested with silting-up rate as a covariate in the same model.
Results Runo capture in diggings The result of the runo capture model for rain events with dierent amounts of rainfall are shown in Table 1. As an assumption of the model, rainfall events of less than 8 mm do not generate runo, and therefore do not generate dierences between densities due to interference. Neither is there interference between the diggings for water once the diggings are ®lled up with water from Table 1 Volume of water received in diggings (as a proportion of digging volume, 795 ml) by position in the plot, during individual rain events >8 mm, for four digging densities (d1±d4, 30, 60, 90 and 120 cm between diggings, respectively) assuming 12 m contributing area above the top row Rainfall (mm) 8 16 24 32 40
Position
Top row Other rows Top row Other rows Top row Other rows Top row Other rows Top row Other rows
Digging density d1
d2
d3
d4
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
1.00 0.84 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
1.00 0.43 1.00 0.87 1.00 1.00 1.00 1.00 1.00 1.00
1.00 0.21 1.00 0.41 1.00 0.62 1.00 0.82 1.00 1.00
rain and runo. Interference between diggings for runo capture only takes place between 8 mm rainfall and the minimum rainfall required to ®ll them, which diers between the digging densities and between the top row and the lower rows, and depends on the assumptions about runo ¯ow from higher up the slope. Our assumption of 12 m probably underestimates the amount of runo generated on the exposed rock outcrop of the lower Shivta formation less than 10 m upslope from the density plots. Our calculations for this amount of runo input in the top row of diggings shows that for the top rows there are no dierences between the digging densities in the amount of water received per rain event. For the other rows, rains of up to 40 mm cause a decrease in the amount of water captured only at the highest digging density. At lower rainfall intensity diggings at lower densities also begin to experience interference. The dierences in water capture between digging densities occur because at high density there is both more intense and more frequent interference. In the winter of 1994±1995 there were only four rain events larger than 8 mm. All other events were small, as is usually the case in the area (Yair and Shachak 1987), and were unlikely to cause dierences in water capture between digging densities. Three early rains (22.4 mm on 5 November 1994, 11.8 mm on 24 November 1994 and 30.9 mm on 3 December 1994) may have caused, according to our model, both dierences between the highest digging density and the rest, and between the top and the other rows at this density. The fourth runoproducing rainfall fell on 4 February 1995 (10.0 mm), which was likely to cause interference at all but the lowest digging density, with increasing intensity at higher digging density. Based on these calculations, we estimated the amounts of water that accumulated in the diggings at the digging density until 5 December and for the whole year (Fig. 1). The assumption of 20 or 30% runo made very little dierence, but assuming a larger capture of the runo ¯ow decreases the dierences between digging densities. Assuming an unrealistically small runo ¯ow of only 1.20 m as input into the ®rst line of diggings has a great eect, but only on the top row of diggings, which now also show interference for runo capture, while the other rows show similar digging density eects. According to the model, we expected that the diggings would only show large dierences by the end of the winter season, mainly because of the small, but runogenerating, event of 4 February 1995. By December, when germination of the annuals was under way, dierences between digging densities were likely to be small. The ®eld soil moisture measurements on the two dates during November 1994 indicated that there were no negative digging density eects on soil moisture in diggings or matrix on either slope. The only dierences we found were between the diggings and the matrix on the south-facing slope (F[1,16] 12.751, P 0.0026 on 10 November and F[1,16] 28.783, P 0.0001 on 15
110
Fig. 1 Estimated volume of water (l) received by diggings during the growing season of 1994±1995 versus digging density (d1±d4, distance between diggings 30, 60, 90 and 120 cm respectively), assuming 12 m input into top row and a digging volume of 795 ml
November 1994). The ®rst measurement was done 5 days after a heavy rainstorm of 22.4 mm, which we assume led to considerable runo ¯ow down the slope. On the second date of soil moisture measurement there was 3.1 mm of rain, below the threshold for runo production. Soil deposition in diggings During the course of the study, the mounds with excavated soil eroded away, ®lling the diggings with on the average 1 cm of soil material, which accounts for 6.69% 0.09% (SEM) of their depth. Although in some of the blocks silting-up of the diggings was somewhat less in plots with high digging density, there were no signi®cant dierences between the digging densities on either slope. The small extent of silting-up led to a negligible reduction in volume of the digging. Assuming an inverted cone shape of the diggings with depth and diameter of 15 cm, 7% ®lling up with 1 cm of soil changed digging volume from 795.22 to 794.98 ml, which is only 0.03%. According to this simple geometric model, it takes 47% silting-up to cause a 10% reduction in the volume of the diggings. As expected, we did not ®nd any eect of these minor changes in digging volume on plant density or biomass production in the diggings. Plant density In the three spring censuses from 1992 to 1995 (Fig. 2), plant density was higher in the diggings than in the matrix for all four digging densities and for both slopes (F[1,474] 807.44, P 0.0001). The mounds hardly
Fig. 2 Mean plant density per patch (number of plants per 100 cm2, SEM) in three kinds of patches (diggings, mounds and matrix) against digging density (d1±d4, 30, 60, 90 and 120 cm) on a north- and a south-facing slope during spring 1992, 1994 and 1995
contained any plants (Fig. 2), and were subsequently left out of the analysis. Throughout the experiment, plant density in both remaining patch types was consistently higher on the south-facing slope than on the north-facing slope (F[1,6] 75.49, P 0.0001 for the location eect). While patch type eects were signi®cant for both slopes, the dierences between the diggings and the matrix were greater on the south- than the north-facing slope (F[1,474] 131.40, P 0.0001 for the interaction of patch type and location). Plant density diered signi®cantly between years for both patch types together for all digging densities (F[2,948] 74.09, P 0.0001). The signi®cant dierences in plant density between diggings and matrix also varied between the years (F[1,948] 39.54, P 0.0001 for the patch type ´ year interaction). This was primarily because plant density in diggings was much lower during the dry year of 1993±1994 than in the others, while matrix patches varied much less. Digging density had a signi®cant eect on plant density over all three years (F[3,6] 15.13, P 0.0033
111
for diggings and matrix together). This was due to the digging density eect in diggings, but not in the matrix (F[2,6] 4.60, P 0.0106 for the patch type ´ digging density interaction). In this respect both locations were similar (F[2,6] 1.07, P 0.3444 for the three-way interaction between location, patch type and density). In diggings, plant communities became consistently less dense with increasing digging density (F[3,6] 19.591, P 0.0017 for diggings only, over the three years together). The signi®cant dierences between years indicate that the digging density eect depended on rainfall, because in the relatively dry year of 1993±1994 with 58.1 mm of rainfall, the negative eect of digging density on plant density in diggings was weaker than in both 1991±1992 and 1994±1995 (with 168.9 and 115.7 mm respectively). Again, both slopes showed this eect similarly (F[6,544] 1.851, P 0.1715 for the interaction between years, locations and digging densities; Fig. 2). During all years, plant density in the diggings decreased from winter to spring, due to seedling mortality. On the north-facing slope this was higher than on the south-facing slope during the two years 1993±1994 and 1994±1995 (F[1,6] 112.64, P 0.0001 and F[1,6] 8.33, P 0.0278, respectively), but not in 1991±1992 (F[1,6] 0.01, P 0.9327). However, mortality rates did not vary according to digging density on either slope during these three years. Therefore, the eects of digging density on plant density in the diggings were essentially the same for both the winter and spring censuses. Total plant density of all diggings and matrix patches per density plot combined decreased with increasing digging density during the ®rst year on the south-facing slope (F[3,6] 5.265, P 0.0406). On the opposite slope there was a downward trend, but this was not signi®cant (F[3,6] 1.431, P 0.3236). In 1994 this was signi®cant on the south-facing slope (F[3,6] 13.524, P 0.0044), but it did not decrease on the north-facing slope (F[3,6] 0.509, P 0.6905). During the third census there was no downward trend on either slope (F[3,6] 3.616, P 0.0845 and F[3,6] 2.964, P 0.1193 for the south- and north-facing slopes, respectively). Plant species The plant communities in the diggings and the matrix in 1992 consisted of 17 species on the north-facing slope and 13 on the south-facing slope (see Appendix 1). Eight species were unique to the north-facing slope, and four to the other slope. In 1994 there were 12 and 27 species on the north- and the south-facing slope, with 2 and 17 unique species, respectively, and in 1995 17 and 21, with 2 and 6 unique species. However, which species was unique to either slope during the three censuses was not predictable. For instance, in 1992 Asteriscus hierochunticus and Euphorbia chamaepeplus were only found on the north-facing slope, and reappeared in 1994
on the south-facing slope only. Other examples are Diplotaxus harra and Hippocrepis unisiliquosa, which were unique to one of the slopes in 1992, but were found on both in 1994 and again only on one in 1995. The inconsistent occurrence of the unique species seems to be a consequence of their rarity, as the commoner species occurred on both slopes. Very few species (Cutandia dichotoma, Emex spinosa, Hordeum spontaneum, Reichardia tingitana and Ononis sicula) were con®ned to one slope throughout the study period. During the 1992 and 1994 censuses, there were fewer than two species per slope that were con®ned to either diggings or matrix patches; in 1995 the diggings contained 2 species out of 17 on the north-facing slope and 6 on the south-facing slope that did not occur in the matrix patches, and these occurred very rarely (Senecio glaucus, on the north-facing slope 4 times and Spergularia diandra 5 times, and Adonis dentata, Bromus rubens, Hippocrepis unisiliquosa, Linaria haelava and Stipa capensis only once on the south-facing slope). All common species occurred in both patch types. On either slope, the commonest species, both in terms of frequency of occurrence and density per patch, were dominant in both patch types. All others occurred in fewer than 30% of the diggings or matrix patches, most in fewer than 10%, with fewer than two individuals on average per patch, where present. On the north slope the common species were, with varying order of abundance over the censuses, Filago desertorum, Herniaria hemistemmon, Plantago coronopus, Picris longirostis, Rostraria cristata and Reboudia pinnata, which were all species of small stature (15 cm). On the south-facing slope the commonest species were Astragalus tribuloides, Calendula arvensis, F. desertorum, Matthiola livida, Plantago coronopus, Picris longirostis, Rostaria cristata, Reboudia pinnata, and Trigonella stellata, of which only R. pinnata, C. arvensis and M. livida attain larger size (>15 cm). Of all individual species, only Rostaria cristata and Plantago coronopus, on the north-facing slope in 1995, displayed the same pattern of decreasing density per digging with increasing digging density (F[3,6] 7.933, P 0.0165 and F[3,6] 12.673, P 0.0052, respectively). Their densities only contributed 28.8% 1.6 (SEM) and 4.7% 0.8, respectively, to total plant density. In 1992 and on the south-facing slope in 1995, none of the species showed the negative digging density eect observed for total plant density. Species richness In the spring censuses, species number per patch was signi®cantly higher in diggings than in the matrix and mounds across all densities and at both locations for all years (F[1,552] 487.137, P 0.0001; Fig. 3). The north-facing slope had higher species number than the south-facing slope (F[1,6] 167.261, P 0.0001). The dierences among years were also highly signi®cant
112
on both slopes (F[3,6] 32.099, P 0.0004), with a marginally signi®cant interaction with location (F[3,6] 4.953, P 0.0461). Digging density also had a signi®cant negative overall eect on species number per patch (F[2,1104] 369.167, P 0.0001). This was based solely on its eect on the diggings in 1995 on the southfacing slope, and in 1992 on the north-facing slope (Fig. 3). In the other years, the decrease in species richness with digging density was not signi®cant, mainly because of the relatively large variability within digging densities, in part between blocks. The negative digging density eect on species number in diggings in 1992 on the north, and in 1995 on the south-facing slope, was not caused directly by digging density, but by the variation in plant density. For all years, plant density per digging (X) accounted for 53% of the variation in species number (Y) on the south slope (Y 3.131 + 0.090 ´ X, F[1,430] 53.419, P 0.0001), but only for 14% on the north slope (Y 1.166 + 0.200 ´ X, F[1,430] 491.717, P 0.0001). In the years when the negative digging density eects were signi®cant, the relationships between plant density and species number were slightly dierent (north slope in 1992: Y 1.063 + 0.197X, F[1,142] 149.753, P 0.0001,
r2 53.1%; south slope in 1995: Y 3.665+ 0.158X, F[1,142] 61.523, P 0.0001, r2 30.2%). With plant density per digging included in the analysis, digging density did not have a signi®cant additional eect on species number (F[3,6] 0.391, P 0.7641 for north in 1992, and F[3,6] 2.577, P 0.1493 for south 1995). Most of the variation in species richness occurred among diggings within digging density plots, not between digging densities. The eect of digging density on total species number in all diggings and matrix patches per plot was not signi®cant, although there was, on both slopes, a downward trend in the ®rst year, from 10.3 0.3 (SEM) species per density plot at low digging density to 5.3 2.0 at the highest density on the north-facing slope, and from 9.3 0.3 to 7.3 0.7 species on the south-facing slope. In 1994 the trend was less clear (from 7.3 0.3 and 16.3 1.3 to 5.7 2.3 and 10.0 0.0 species for the north and south slopes respectively). During the last year, total species number did not decrease at all with increasing digging density. Plant productivity Plant productivity in the sample patches harvested in 1995 behaved similarly to plant density. It was signi®cantly higher in diggings than in matrices on both slopes (F[1,264] 150.005, P 0.0001), and there was a sig-
Fig. 3 Mean number of species per patch (SEM) against digging density (d1±d4, 30, 60, 90 and 120 cm) on a north- and a south-facing slope in spring 1992, 1994 and 1995
Fig. 4 Mean annual biomass production (g dry mass per 100 cm2, SEM) in diggings and matrix against digging density (d1±d4, 30, 60, 90 and 120 cm) during spring 1995 on a north- and a south-facing slope (total rainfall 115.70 mm)
113
ni®cant negative eect of digging density (F[3,6] 34.358, P 0.0004), but only in the diggings, not in the matrix patches (Fig. 4). Similarly to plant density, productivity was higher on the south-facing than on the north-facing slope (F[1,6] 256.487, P 0.0001). Unlike species richness, productivity was signi®cantly aected by digging density itself, without mediation by plant density (F[3,6] 31.039, P 0.0005 for the north-facing slope, and F[3,6] 12.395, P 0.0065 for the south-facing slope). Only on the south-facing slope was part of the response to digging density of productivity in diggings also based on the eect of plant density on productivity (F[1,59] 8.004, P 0.0064). On the north-facing slope, productivity of the vegetation in diggings varied independently of plant density. As was also found for plant density, biomass production was not signi®cantly aected by the level of ®lling-up of the diggings on either slope.
between diggings when the distances between diggings are low. If there are also dierences between the top and all other rows further down the slope within the density plots, especially at high digging densities, interception of runo ¯ow may be responsible. In order to dierentiate between interception of runo and seed capture, we compared biomass production in 1995 of individual harvested diggings within the density plots between the top row of diggings, those in the bottom row and those in the middle rows. At the lowest digging density there were no dierences between the top and the bottom rows, but they were signi®cant at higher digging densities (Fig. 6). The position eects were signi®cant for both slopes (F[1,24] 4.923, P 0.0362 on the north-facing slope, and F[1,16] 52.248, P 0.0001 on the south-facing slope). Only on the south slope did the dierences in position eect between digging densities result in a signi®cant
Seed densities Estimated densities of seeds that accumulated in the diggings from outside after the growing season of 1994± 1995 were aected by digging density only on the northfacing slope, but not on the opposite one (F[3,16] 4.434, P 0.0189 for digging density on both slopes together, with F[3,16] 3.673, P 0.0347 for its interaction with slope dierences; Fig. 5). The lack of a density eect on the south-facing slope coincided with very high productivity values during the previous winter, but, in contrast to the north-facing slope, with low average seed numbers. Position on the slope The negative eect of digging density on plant density and biomass production in diggings implies interference
Fig. 5 Seeds trapped in diggings from outside during summer 1995 (number per patch, SEM, estimated in germination tests), versus digging density (d1±d4, 30, 60, 90 and 120 cm) on two slopes
Fig. 6 Mean productivity (g dry biomass per 100 cm2, SEM) in diggings against digging density (d1±d4, 30, 60, 90 and 120 cm) and position on the slope during spring 1995 on a north- and a southfacing slope
114
interaction between position and digging density (F[3,16] 5.150, P 0.0111). However, even the bottom row diggings, while lower than the top row, had higher biomass production than the middle rows. This is not consistent with the expectations of our runo capture model, which speci®es dierences between the top and all other rows, if runo is responsible for the negative digging density eect.
Discussion Our results provide evidence that small soil pits interfere with each other because plant density, species number and biomass production of the annual plant communities in the diggings decrease as digging density increases. Although there are some dierences, this phenomenon was found to occur on both a north- and a south-facing slope of the same watershed. In general, the negative eects of digging density occur at the level of the plant communities through changes in density of all individual species together. We found no evidence in favor of runo down the slope as the factor responsible for the negative eect of digging density on plant density and species number in the diggings, nor for biomass production in 1995. Therefore, we suggest that the main cause of the digging density eect is interference in the capture of seeds, causing lower plant density in diggings at high digging density. This negative digging density eect occurs in addition to the dierences between diggings and undisturbed matrix, which depend on enhanced capture of both runo water and seeds in diggings relative to the matrix (Shachak et al. 1991; Boeken et al. 1995). The main conclusion from the runo model is that interception of runo ¯ow down the slope should cause dierences between the digging densities as well as between the top row of diggings and the ones further down the slope, at higher digging densities. It also showed that the dierences among the digging densities in the amounts of runo water captured in 1994±1995 are small, at least during the ®rst months of the season. The position eect, according to the model, is mainly due to the large runo input from the Shivta rock outcrop upslope, and the small size of the diggings. The vegetation responses did dier between the digging densities, but not between the top row and the others, as should be the case if runo interference were the cause of the observed dierences. For instance, the dierences we found between the rows in biomass production were greater in the bottom row, which cannot be accounted for by runo interception. In reality, over¯ow of excess runo water may follow runnels, and may not be captured on the way down. In that case, there should only be differences between the top and lower rows in all digging densities, but not between the digging densities themselves, which is not in accordance with any of the observed vegetation responses.
The soil moisture measurements, showing no dierences between diggings at dierent densities, also indicate that runo capture is not the main factor causing interference between diggings. These ®ndings suggest that runo capture in all diggings is sucient to cause high soil moisture, regardless of the density of the diggings. The independence of local digging density is consistent with the conclusion of Boeken et al. (1995) that the conditions in diggings are better than in the undisturbed matrix due to runo capture, but that even the smallest amounts of runo ¯owing down the slope from the rock outcrops are sucient. In addition, even if there are dierences in the amounts of water captured, these may not determine variation in recruitment, as long as soil moisture is above a minimum required for maximum seed germination and seedling establishment. We tend to favour the alternative hypothesis, that the digging density eect on plant density is caused not by runo capture, but by seed capture. This is supported by the seed density estimates, which were based on capture of seeds dispersed after the highly productive growing season of 1994-1995. Since seed capture was assessed until the beginning of the rainy season, the estimates are independent of runo capture. Interference in seed capture implies seed limitation of plant density, as opposed to site-limited recruitment (Crawley 1986), which in annual desert plant communities is strongly correlated with water limitation. In some cases, seed-limitation may be more critical than well-known water-limitation (Went 1948, 1949; Noy-Meir 1973; Evenari et al. 1982; Gutterman 1993) in determining the dynamics of desert annual plant communities. The eect of between-digging interference is primarily on plant density and biomass production, while its eect on species number is due to its covariance with plant density. In contrast to interference between diggings, which occurs at a scale including several diggings, in areas diering in digging density, species number varies among individual diggings, due to its well-known relationship with plant density (Peet 1974). The more individual plants there are in a digging, the greater the chance of having more species, but this does not directly depend on digging density. This indirect relation between digging density and species number also causes only a very weak decrease in overall species diversity per density plot with increasing digging density. Biomass production increases as the plant community becomes denser, but in contrast with species number, this accounts for only a small part of the variation in biomass production. More plants means more productivity, but only at the same level of resources; given sucient available resources, communities with fewer individuals can have greater biomass than denser communities. Although the observed pattern of productivity is inconsistent with our expectations of the eects of runo interference, dierences in quantities of water per digging are undoubtedly an additional factor in determining plant productivity. Further study is needed to
115
determine its precise mechanism and its interaction with seed capture. Digging density does not have a negative eect on the communities outside the diggings, on the undisturbed soil matrix. This con®rms previous observations on the eects of soil disturbance: pits are good at capturing runo water and seeds (Gutterman 1993, 1997; Shachak et al. 1991; Boeken and Shachak 1994; Boeken et al. 1995), while in the matrix the net accumulation of resources and seeds is low because runo is generated there (as well as on exposed rock surfaces; Yair and Shachak 1987; Shachak et al. 1991; Boeken et al. 1995). This, together with exposure to wind, and the relative ease for ants to remove seeds from undisturbed soil (Gutterman 1993; Boeken and Shachak 1994), makes the soil matrix a rather unfavorable place for plant establishment and growth, where both seeds and resources are limited. The eects of digging density on the annual plant communities in diggings are aected by year-to-year dierences, and by local dierences between slopes, as re¯ected in the statistically signi®cant interaction between digging density, years and slopes. Years dier primarily in totals and temporal distributions of intensity and amount of rainfall (Gutterman 1995; Gutterman and Evenari 1994). In our study it seems that interference among diggings is weaker during years of low rainfall (as in 1991±1992). In wet years (such as 1994±1995), the eects in the diggings were greater. The dierences between the south- and the northfacing slopes used in the experiment can not be explained adequately. This illustrates that there are more factors involved in determining diversity and productivity of annual plant communities. Speci®cally, local dierences are not only attributable to dierence in direction and exposition, but also to soil depth and chemistry (Friedman and Orshan 1975), exposed rock surface area (Shachak and Brand 1991), and shrub cover (Boeken et al. 1995). Slope direction should make north-facing slopes cooler than south-facing slopes, as Boeken and Gutterman (1990) demonstrated on the same slopes, but this may not be relevant for sunken diggings (Gutterman 1997a). Especially enigmatic are the greater seed numbers on the north-facing slope in the 1995 season, in spite of higher plant densities on the opposite slope. These considerations lead us to propose a mechanism of patch interference. The observed eects of digging density on seed capture, plant density, species number and productivity suggest that diggings interfere with each other in the capture of seeds. Boeken and Shachak (1994) and Boeken et al. (1995) also showed that seed capture is an important factor determining annual plant dynamics in the Negev desert. The evidence stems from very low seed availability in disturbed compared with undisturbed soil (Boeken and Shachak 1994) and from a strong correlation of species richness in natural porcupine diggings (Boeken et al. 1995) with the presence of species-rich patches associated with shrubs (Weinstein 1975). We propose that diggings capture seeds deposited
in their vicinity during ``phase I'' dispersal from the seed rain, by functioning as a sink for ``phase II'' dispersal along the ground (Watkinson 1978; Chambers and Macmahon 1994). The fact that nearly all species in the diggings possess small (