Ecogeography and Demography of Cicer Judaicum Boiss., a Wild Annual Relative of Domesticated Chickpea

Published online April 25, 2006

Ecogeography and Demography of Cicer judaicum Boiss., a Wild Annual Relative of Domesticated Chickpea

Roi Ben-David, Simcha Lev-Yadun, Canan Can, and Shahal Abbo*

ABSTRACT domesticated chickpea compared with other cool season Chickpea (Cicer arietinum L.) is a major pulse crop in the Indian grain legumes (e.g., Abbo et al., 2003; Siddique et al., subcontinent and other world regions and is characterized by narrow 1999; Turner et al., 2001). Indeed, in their review of adaptation relative to other cool season food legumes. Comparative pulse crops adaptation in the Canadian and U.S. northern ecophysiology employing closely related wild species is a powerful Great Plains, Miller et al. (2002) noted that chickpea has tool to broaden the understanding of the genetic and physiological an “unusual response to growth season rainfall” and that basis of crop adaptation. However, meager data are available on the pea and lentils had broader adaptation compared with ecological preferences of annual wild Cicer species. Moreover, the chickpea. Miller et al. (2002) also stressed that “compar- geographic range, its size, shape, and boundaries as well as its in- ative adaptation among pulse crops remains poorly un- ternal structure have never been studied for any of the wild Cicer taxa, derstood.” Regardless of the evolutionary and agronomic thereby limiting our understanding of Cicer biology. Accordingly, this work focused on Israeli C. judaicum Boiss. a wild annual relative of reasons for the failure of chickpea to occupy a major chickpea. We defined the range of the species across the Mediterra- niche among the autumn sown crops of temperate lati- nean zone of Israel, characterized the ecogeographical profile of its tudes, understanding the genetic and physiological basis habitats and studied two populated sites at the macro- and microsite of adaptation is fundamental for future improvement of levels in terms of plant density, frequency, and niche physical char- chickpea and other crop plants (e.g., Evans, 1993). acteristics. Throughout the survey area the species is mostly confined The closely related wild relatives of crop plants pro- to stony and rocky niches where competition with more aggressive vide excellent material to study the potential adaptation annuals is small. This habitat preference dictates a patchy distribution spectrum of the respective crops (Evans, 1993). Eco- pattern at all levels, from local niche to the region and beyond. physiology and comparative physiology are key components in such an approach (therein), provided there HICKPEA is a grain legume associated with the ini- is a wide range of germplasm lines representing a suf- Ctiation of agriculture in the Neolithic Near East ficiently wide ecological spectrum. In the case of Cicer, (Lev-Yadun et al., 2000; Zohary and Hopf, 2000). An- however, the number of independent wild annual acces- nual world production of chickpea amounts to approx- sions found in genebanks is relatively small (Berger et al., imately 8 3 106 Mg, a lion’s share of it (94%) in devel- 2003). The data available from the annual Cicer world oping countries, mainly in the Indian subcontinent collection indicate that they provide incomplete coverage (FAO, 2006). Large population sectors in India and its of the geographical distribution of the taxa and thus prob- neighboring countries, many of them vegetarian, depend ably also of their genetic diversity (Berger et al., 2003). on chickpea as a main dietary protein source, while in As part of our ongoing project aimed at broadening developed countries it is considered a health food (e.g., the understanding of the genetic and physiological basis Abbo et al., 2005). of chickpea adaptation, we focused in this study on C. In a recent review on the evolution of domesticated judaicum, an annual wild taxon and a member of the chickpea (Abbo et al., 2003), the authors suggested that tertiary genepool of the cultigen (Ladizinsky and Adler, the restricted distribution of C. reticulatum Ladiz. (the 1976). This species is cross-compatible with C. bijugum immediate wild progenitor) poses severe limitations on Rech. f., which grows in higher latitudes (Ladizinsky and the adaptation of the crop. These authors hypothesized Adler, 1976), making both taxa suitable parents for seg- that the narrow adaptation of C. reticulatum (as deduced regating populations to study the genetic basis of adap- from its restricted natural distribution) represents (or tation (Abbo et al., 2003; Berger et al., 2003). The stems from) limited genetic variation in key adaptive loci specific aims of this work were as follows: first, to define (e.g., temperature and day-length response genes and the geographic range of C. judaicum across the Medi- edaphic requirements). Further, this narrow genetic vari- terranean zones of Israel and characterize this range by ation could partly account for the adaptive limitations of its size, shape, boundaries and internal structure; second, to identify ecogeographic patterns in the distribution of R. Ben-David and S. Abbo, The Robert H. Smith Institute of Plant C. judaicum by using Geographic Information Systems Science and Genetics in Agriculture, The Hebrew University of (GIS) and multivariate analyses; and third, to charac- Jerusalem, P.O. Box 12, Rehovot 76100, Israel; S. Lev-Yadun, terize the local spatial pattern of C. judaicum plants, Reproduced from Crop Science. Published by Crop Science Society of America. All copyrights reserved. Department of Biology, Faculty of Science and Science Education, their density, frequency, and preferred microhabitats by University of Haifa-Oranim, Tivon 36006, Israel; C. Can, Department focusing on two selected populations. of Biology, University of Gaziantep, 27310, Turkey. Received 12 Oct. 2005. *Corresponding author ([email protected]).

Published in Crop Sci. 46:1360–1370 (2006). MATERIALS AND METHODS Plant Genetic Resources Survey Procedure doi:10.2135/cropsci2005.10-0351 ª Crop Science Society of America A field survey was held during the growth seasons 677 S. Segoe Rd., Madison, WI 53711 USA (November–April) of 2002, 2003, 2004, and 2005 searching 1360 BEN-DAVID ET AL.: ECOLOGY OF WILD CICER JUDAICUM 1361

for C. judaicum populations across 236 sites in central and PCA Analysis northern Israel, excluding most of the West Bank area. First, localities that were documented in the herbarium of the Hebrew Principal component analysis (PCA) of the continuous University of Jerusalem or in the “ROTEM” database (BioGis, ecogeographical variables was used to identify a smaller num- 2002) were visited. Second, typical “like-habitats” were sampled ber of principal components to account for most of the using the “search image” concept (Engels et al., 1995). To es- ecogeographical variance among 52 sites populated with C. tablish presence or absence patterns, ecogeographic data was judaicum. PCA was based on the correlation matrix and collected for each site irrespectively of whether C. judaicum was presented as biplot ordinations of populated sites (PC scores) found. In each site, latitude, longitude, and altitude were de- and ecogeographical variables (PC factor loading). Actual termined by GPS, and location descriptors (geographical region, populated sites were depicted in the ordination by plotting road or settlement name, proximity to prominent land marks) sites’ principal component scores using a unique symbol for and site physical characteristics (habitat type, slope, aspect, each cluster of sites. Three components were extracted by and precise location of target species plants at the site, if found) Eigenvales .1 to ensure meaningful implementation of the were recorded. Ecogeographical data were extracted from the data by each factor. PCA and hierarchical clustering was Geographic Information System (GIS) center of the Hebrew applied by SPSS v.12 (SPSS, 2002). University of Jerusalem and from maps of the Israel meteo- rological service (for annual and monthly rainfall, monthly Ecological Assay minimum and maximum temperature, soil type, lithology, associated species, vegetation cover, and distance to nearer The ecological assay, aimed at characterizing the local con- settlement, road, or reserve). To ensure independence between ditions at the populated site level, was preformed at Nahal observations and avoid autocorrelations, a minimal distance of A’naba and Nahal Mea’ra (Fig. 1) on four different dates 2 km between sites was kept (Legendre, 1993; Guisan and (March 2002 and three dates during 2003 season: early Jan- Zimmerman, 2000). On the basis of the presence–absence data, uary, end of February and early April). In addition, two long a “dot map” was constructed by the GIS software ArcView3.2 transects, extending from south to north and from west to east, (ESRI, 2002). were undertaken in each site during the 2003 season to obtain a rough estimate of the spatial size of the two populations. Multivariate Analysis The N. A’naba site is located southeast of Modii’n town and consists of four slopes: southeast, west, northwest and east The following 11 ecogeographical variables were used in facing slopes. The N. Mea’ra site is located at Judea foothills all multivariate techniques: longitude (Ln), latitude (Lt), alti- and consists of two contrasting, south and north facing slopes. tude (Al), annual rainfall (Rn-A), mean and minimum tem- Both sites are under seasonal grazing, cattle at N. A’naba and perature in January, representing peak winter conditions sheep at N. Mea’ra. (mean and minimum winter temperature- Tmean-W, Tmin- Demographic surveys in both sites were performed along W), mean and maximum temperature in August represent- random transects across the different slopes (2–4 transects per ing peak summer conditions (mean and maximum summer slope) aiming to maximize microsite topographic variability. temperature—Tmean-S, Tmax-S), slope direction relative to Transects position relative to slope direction was random. the north (S-D), slope rate in percentage (S-R), and distance to Each transect (50 3 0.5 m) consisted of 25 sampling units of sea (Dis-S). For the implementation of discriminant analysis 1m2 each. Two ecological parameters were measured in each and hierarchical clustering analyses, the ecogeographical data sampling unit, namely, frequency and plant density. Frequency were standardized by dividing each descriptor by its range (presence, regardless of the number of individuals, or absence (maximum–minimum). of C. judaicum from the sampling units) was measured as “Rooted (plant) Frequency” (Greig-Smith, 1964) on the four Discriminant Analysis visiting dates. Average frequency of eachp transect was transformed by the transformation arcsin x (0 , x , 1) The availability of presence–absence data, from the 236 sur- (Sokal and Rohlf, 1997). Plant density (number of plants in vey sites, allowed the use of a group discriminant technique each sampling unit) was measured during the 2003 season only. aimed to optimize the prediction model performance (Robertson Average plant density of each transect was transformed to et al., 2003). This was made deliberately to optimize the pre- p (x 1 1) for analysis of variance and Tukey Kremer test (after diction model performance (Robertson et al., 2003). Canoni- Noy-Meir et al., 1991b). A third ecological parameter, the cal discriminant analysis was performed with absent–present distribution pattern (I value), was calculated for each transect as the initial classification criterion. The prediction model out of the plant density data, with I 5 (variance of sampling group- discrimination technique was based on variables that units within the transect)/(average plant density in the may affect plant growth directly (Rn-A,Tmean-W, Tmin-W, transect) (Kershaw, 1973). Tmean-S, and Tmax-S) and indirect variables mediating their To assess the effect of site and sampling date on the three effect via the direct variables (Ln, Lt, Al, S-D, S-R, and Dis-S) ecological variables (frequency, plant density, and distribution (defined by Austin, 1985). The relatively limited geographical pattern), analysis of variance (ANOVA) was implemented. An range of the survey justifies the utilization of indirect vari- Reproduced from Crop Science. Published by Crop Science Society of America. All copyrights reserved. unbalanced factorial model was employed with Y as the ables (Guisan and Zimmerman, 2000). Subsequently, the ijk score of the kth transect sampled at the jth sampling date effects of the qualitative variables of soil type and lithology from the ith site for each of the three ecological variables, by were assessed through a GIS query. Discriminant analysis was the formula: preformed by the JMP 5.0 statistical package (SAS Insti-

tute, 2001). Yijk 5m1aI 1bj 1aI 3bj 1 eijk where m is the general mean, a is the site factor with fixed Hierarchical Clustering I effect [i 5 1 (N. A’naba), i 5 2 (N. Mea’ra)], bj is the sampling Ward’s hierarchical clustering (Ward, 1963) was used to date factor with fixed effect [j 5 1 (early January), j 5 2 (end of

identify discrete groups of C. judaicum populated sites, after February), j 5 3 (early April)], aI 3bj is the effect of the

Delacy et al. (1996) and Berger and Speijers (2003). interaction between the jth sampling date and ith site, and eijk 1362 CROP SCIENCE, VOL. 46, MAY–JUNE 2006

Fig. 1. Distribution map of C. judaicum sites across Israel. Collection sites of C. judaicum specimen deposited at the Hebrew University of Jerusalem Herbarium (s), sites found in this work (E).

is the effect of the kth transect in the ijth aI 3bj combina- spective. The present geographical range found in this tion (k 5 1ton). survey is in accord with the historical range with the On the basis of the frequency data, a total of 160 pop- exception of an herbarium specimen collection around ulated patches were detected in both sites, 99 at the N. A’naba Metula (E358349 N338169) in the upper Galilee (Fig. 1). site and 61 at the N. Mea’ra site. This nonbiased sample of In our survey, no population was found north of the 160 patches was used for the assessment of descriptive statistics of patch characteristics at both sites and comparisons Nazareth mountains (E358159 N328459) anywhere in the between sites (Student’s t test). lower and upper Galilee, despite intense efforts. General habitat classification was recorded for each sam- The Israeli C. judaicum populated sites fall into two pling unit (644 in total). Each of the 25 sampling units in each main groups, northern and southern. This finding partly transect was assigned to 1 of 10 microhabitat classes: rocky, reflects the poor sampling within the Palestinian Au- stony, exposed, grove, and their six combinations. Composi- thority area of Samaria and Judea and the dense human tional habitat diversity was measured by calculating Shannon infrastructure in the Sharon plain and lower parts of diversity index (SDI) (Magurran, 1988) for each transect, Samaria, which severely disturbed known C. judaicum natural habitats. The main geographic regions in Israel H952Rpilnpi in which C. judaicum was absent include Lakhish, the 2 where pi is the proportion of m sampling unit at each transect costal plain, Mt. Gilboa, western Galilee, upper Galilee, classified to the ith class. and the Golan heights. To examine the association between frequency data and habitat type, a x2 test for discrete variables was performed. Ecogeographical Characterization of C. judaicum Sites RESULTS The altitude of C. judaicum sites ranges from 41 m Survey above sea level (a.s.l.) in Nahal Narbata to 749 m a.s.l. in Reproduced from Crop Science. Published by Crop Science Society ofOverall, America. All copyrights reserved. 52 C. judaicum populations were found in Mevo Beitar. The average altitude across all the popu- this survey from Haela Valley (Aviezer) in the south to lated sites is 322 6 15 m. Each year, a typical C. judaicum Nazareth Mountains in the north (Fig. 1). The northern site gets between 55 and 65 rainy days (.0.01 mm population (Hasolelim) and the southern population rainfall). The mean annual rainfall (Rn-A) ranges from (Aviezer) are only 120 km apart. Hasolelim is also the 473 mm in Aviezer to 665 mm in Mey A’mi, with an eastern most populated site and is located 30 km from average of 564 6 8 mm across sites. The main rainfall West Carmel, the western most population (two km off share (73%) occurs during the vegetative phase of the the Mediterranean seashore). Incorporation of her- C. judaicum plants (mid November to mid February), barium data into the map (Fig. 1) enabled us to define with the rest occurring either as germinating showers the range of the species in Israel from historical per- (4% of the rainfall) or during flowering and podding BEN-DAVID ET AL.: ECOLOGY OF WILD CICER JUDAICUM 1363

Table 1. Mean 6 SE and ranges of seasonal rainfall and tem- Discriminant Analysis peratures (during the growth season) for sites with C. judaicum populations in Israel. The discrimination between sites designated “C. Month of Range of maximum Range of minimum judaicum present” and “C. judaicum absent” was growth season temperature† temperature† Rn-M found significant by the Wilks’ Lambda test [P(F) # °Cmm0.0001]. The output data of C. judaicum absence sites November 21.0–24.5 12.3–14.4 84 6 1 that were classified into the C. judaicum present group December 15.9–20.0 8.0–11.8 121 6 4 can be used to highlight areas of unpopulated sites with January 14.0–20.0 0.9–3.0 139 6 1 February 14.0–18.0 6.2–9.2 101 6 1 the potential to host C. judaicum (Fig. 2). The western March 17.0–21.0 7.7–10.4 74 6 1 lower Galilee (6 sites), Mt Carmel (3), west Samaria (4), April – – 25 6 1 and south Judea mountains (24) are such areas. In this 6 May – – 3 0.3 respect, the Galilee region is unique, since no population † missing data for April and May. was found there so far. This region is between Hasolelim (northern most population of the current survey) and Metula, where the northern most specimen of the stages (26%, see Table 1). In the winter, the minimum Hebrew University Herbarium was collected by the temperature at C. judaicum sites is relatively high, late M. Zohary in 1965. ranging between 7 and 9.58C. In November, at the beginning of the growth season, the minimum–maximum Ecogeographic Clusters and PCA Analysis range is 13.4 to 22.68C, and becoming colder in De- cember (9.52–18.68C) and January (2.4–178C). At the The hierarchical clustering procedure assembled the end of the winter, temperatures are relatively mild, 52 C. judaicum sites into three distinct clusters. This ranging between 7.8 and 17.68C (Table 1.). three-cluster solution accounts for 87% of the variation The survey data do not show any habitat preference in the data set. Cluster no. 2 is the biggest including 29 of the species to a certain slope aspect. Populations are located at southern (37% of the populations), western (32.6%), northern (24%), and eastern slopes (6.5%). This distribution pattern of the populated sites is not significantly different from the division of slope aspects among all survey sites. The low frequency of east facing slopes is dictated by the topographic structure of the Judea and Samaria Mountain ranges, which dominates the survey areas. The slope rate of the populated sites is moderate to medium, though some are completely flat or steep, both at about 10% frequency. At 89.2% of the C. judaicum sites, the soil is defined as Red Mediterranean, which consists of Terra Rossa and brown rendzina soils (21 and 20 of the population sites, respectively). These two types of soils are well aerated, have good water percolation and high field ca- pacity (Ravikovitch, 1969). In accordance with the country’s lithology, Terra Rossa soils are usually formed on hard limestone and dolomite bedrock (90% of Terra Rossa soils), while the brown rendzina sites mostly rest on chalk (76% of them).

Vegetation Characteristics of C. judaicum Sites About 40% of the populated sites are dispersed in planted forests (mostly pine) and 30% in natural habitats, both much disturbed by human activity. Among the vegetation forms in the populated sites, batha, dense Reproduced from Crop Science. Published by Crop Scienceand Society of America. All copyrights reserved. open garigue, maquis, and park vegetation were recorded. In the natural habitats the common perennial species found were the following: Quercus calliprinos Webb., Quercus ithaburensis Decaisne, Pistacia lentiscus L., Ceratonia siliqua L., Rhamnus lycioides L. subsp. Fig. 2. Prediction of sites with a potential to host C. judaicum. graeca (Boiss. et Reuter) Tutin, Sarcopoterium spinosum Analysis is based on ecogeographical data from 236 survey sites. (L.) Spach, Calicotome villosa (Poiret) Link and The prediction model group discrimination techniques used records of 9 quantitative variables. In a succeeding stage the qualitative Majorana syriaca (L.) Rafin. Common annuals at the variables of soil type and lithology were applied as selecting factors sites were Trifolium spp., Medicago spp., Vicia spp., and by the GIS software. Sites where C. judaicum may be found (j), Onobrychis spp. sites where the species is unlikely to grow (s). 1364 CROP SCIENCE, VOL. 46, MAY–JUNE 2006

Fig. 3. Principal components analysis (based on the correlation matrix) of the 52 C. judaicum sites from Israel, based on ecogeographic variables (indicated in materials and methods). Biplot vectors indicate strength and direction of factor loadings for PC1 and PC2. Points are sites scores with cluster membership (see Table 3) illustrated with different symbols.

populated site, while Clusters 1 and 3 contain 13 and 9 The position of C. judaicum sites marked by clusters populated sites, respectively. on the plane of the first two PC axes highlights the PCA extracted three principal components (Eigen- ecogeographic variation between and within clusters. values . 1), accounting collectively for 85.5% of the Cluster 1 scores mostly concentrated at the positive part variance. Principal component 1 [PC1] (x-axis, Fig. 3, of the PC2. It consists of northern sites from the lower Table 2) explains 46.2% of the dataset variation. This Galilee, Mt Carmel, and western Samaria. These sites axis is loaded positively with altitude, distance from the characterized mainly by high rainfall and relatively short sea, longitude, mean annual rainfall, and slope. PC1 is distance from the Mediterranean Sea. Cluster 2, the also negatively loaded with winter and summer temper- largest cluster includes sites from the Shfela and the ature variables. PC 2 (y-axis, Fig. 3, Table 2) explains Judea foothills. These sites scores mostly concentrated 29.5% of the dataset variation, and is positively loaded at the negative part of the PCA chart. It consists of with latitude and mean annual rainfall. PC2 is also western sites characterized mainly by short distance negatively loaded with maximum summer temperature from the Mediterranean Sea, low altitude as well as and distance from the sea. The angle to north variable low rainfall, relatively high winter temperature, and exhibits weak factor loadings for PC1 and PC2. In ad- relatively intermediate-to-high summer temperature. dition, PC3 positive factor loading is dominated almost Cluster 2, holds the widest ecogeographic range as ex- exclusively by the angle to north variable, but is asso- pressed by site score distribution (Fig. 3). The nine sites ciated with only 9.9% of the dataset variation with a of Cluster 3, originating from the Judean Mountains, marginal Eigenvalue (Table 2). are plotted along the positive part of PC1 reflecting southern location, high seasonal rainfall, and low winter and summer temperatures (Fig. 3, Table 3). Table 2. Factor loadings for principal components with Eigen- values . 1 and % of variance explained by each component. Population Demography Component Most of the C. judaicum populations were small and Variable 1 2 3 plants occurred sporadically in space, as little groups Longitude 0.62 0.38 0.28 or as individuals. At the microsite level, C. judaicum Reproduced from Crop Science. Published by CropLatitude Science Society of America. All copyrights reserved. 20.27 0.92 0.03 plants occupied different habitats such as open grass- Annual rainfall 0.47 0.72 20.03 Mean January temperature 20.97 0.20 20.07 land, rocky crevices, ancient dry stone walls, and rub- Maximum August temperature 20.91 20.30 0.10 ble mounds. Minimum August temperature 20.17 20.93 0.19 The estimated spatial size of N. A’naba population, Mean January temperature 20.99 20.09 20.03 Altitude 0.96 20.26 0.04 on the basis of the two frequency transects, was about Distance to sea 0.69 20.70 0.13 1280 3 903 m (estimated area of 115.5 3 104 m2). In Slope 0.46 0.07 20.72 practice, there was a continuum of C. judaicum plants Angle to north 0.13 0.40 0.64 Eigenvalues 5.07 3.24 1.08 and the population was delimited to the north and east Percentage of variance 46.16 29.45 9.85 by the newly developed area of Moddi’n town (observa- Cumulative percentage 46.16 75.60 85.46 tions suggests that the original range was much wider). BEN-DAVID ET AL.: ECOLOGY OF WILD CICER JUDAICUM 1365

Table 3. The C. judaicum populated sites across Israel merged into three ecogeographic clusters. For each cluster means 6 SE of ecogeographic descriptors and number of populated sites are listed. The hierarchical clustering analysis was preformed on the 52 populated sites using 11 ecogeographic variables (indicated in materials and methods) after Ward (1963). To compare means, Tukey Kramer test was performed. In each variable, means labeled with different letters differ significantly at a 5 0.05. Cluster Number of Mean annual Mean winter Minimum winter Mean summer Maximum summer Altitude above Distance to no. populated sites rainfall (mm) temperature temperature temperature temperature† sea level sea °C mkm 1 13 612 6 11a 11.3 6 0.2a 2.4 6 0.1a 25.2 6 0.1a 31.4 6 1.3 155 6 27a 14.2 6 1.6a 2 29 528 6 8b 10.9 6 0.1a 2.6 6 0.1a 25.7 6 0.1b 34.4 6 0.4 247 6 18b 28.4 6 1.1b 3 9 617 6 14a 8.7 6 0.2b 1.3 6 0.1b 24.2 6 0.1c 32.8 6 0.6 630 6 32c 39.3 6 1.9c † Tukey Kramer test was not implemented since the assumption of homogeneity between clusters was rejected (Bartlet test P(F) # .05).

The estimated size of the N. Mea’ra population was Populated Patch Descriptors smaller, being 784 3 110 m (estimated areaof 8.6 3 104 m2). The patchy distribution of C. judaicum population, as Population Ecological Parameters seen in both sites, stressed the need to focus on the populated patch level. Some of the patch descriptors are In all three ecological descriptors, there were differ- significantly different in the two sites (Table 4). Patches ences between sites (Table 4). I values of both sites are bigger and thicker at N. A’naba than at N. Mea’ra, yielded a ratio higher than 1, pointing to a patchy dis- as evident from the number of plants, estimated patch tribution of the two populations. I ratio was significantly area, and plant density within the patch. Average dis- higher at N. A’naba than at N. Mea’ra (5.6 and 2.9, tance between populated patches of 8.2 m was cal- respectively, see Table 4), indicating a more aggregated culated for both sites. Distribution pattern values of pattern. This patchy pattern is in accord with the habitat I . 1, I 5 1, and I , 1 represent patchy, random, and heterogeneity of the two sites (Fig. 4). regular distribution pattern, respectively. The secondary The frequency and density values are significantly aggregation found within patches at N. A’naba (I 5 2.3), higher in N. A’naba population than in the N. Mea’ra could be a possible indication for microheterogeneity population (Table 4). At N. A’naba, C. judaicum plants at the patch level. Another indication for such second- were found in 228 out of a total of 601 sampling units ary aggregation is the high variance within the patch (frequency of 0.37). At N. Mea’ra, the species was ob- detected at N. A’naba population (Table 4). Unlike the served only in 97 out of 575 sampling units (frequency secondary aggregation at N. A’naba site, the distribu- of 0.14). The density parameter was tested across three tion pattern of C. judaicum plants at N. Mea’ra within sampling dates during the 2003 season. Significant patches was fairly regular. changes in density values were observed in both sites during the growth season (comparing early January with Microhabitat Characterization the end of February and early April, Fig. 5). Unlike the seasonal changes in the density values, the measured Habitats with stony characteristics were the most frequency values in January did not differ from February common in the tested sites. The majority of the sampling or April records. units were counted as one of the following: stony, stony–

Table 4. Demographic parameters of C. judaicum populations at N. A’naba and N. Mea’ra, and distribution of plants in the populated patches therein. Data are mean 6 SE of the distribution pattern within site and within patch (I), frequency of C. judaicum plants in each site and density values within site and patches, estimate patch area, distance between patches, number of plants per patch and variance of number of plants within spot. All patch descriptors data except for distribution pattern are based on 99 patches sampled at N. A’naba and 61 patches at N. Mea’ra. Probability of F and t values indicates the significance of differences between the two sites when applying ANOVA to the listed parameters. Site mean 6 S.E Demographic parameter N. A’naba N. Mea’ra Probability of F value for site factor (R2)† Distribution pattern (I value) 5.61 6 0.7 2.9 6 0.6 0.005 0.32 Frequency‡ 0.37 0.14 0.0001 0.32 Plant density (plants/m2) 1.8 6 0.1 1.4 6 0.1 0.0003 0.7

Reproduced from Crop Science. Published by Crop SciencePopulated Society of America. All copyrights reserved. patch descriptors Probability of t value for site factor Mean 6 S.E of both sites Estimated patch area (m2)§ 1.45 6 18.4 9.6 6 1.9 0.01 – Distance between patches (m) – – NS 8.2 6 0.7 Number of plants 13.1 6 1.6 4.8 6 2 0.0004# – Plant density (plants\m2) 3.5 6 0.2 2.2 6 0.3 0.0004# – Distribution pattern (I value)¶ 2.3 6 0.4 0.8 6 0.7 0.002# – Variance of number of plants within spot 16 6 3.5 2.39 6 6.5 0.0014 – † R-square goodness of fit coefficient of the factorial model. For detailed model see material and methods. ‡ The frequency values were transformed before analysis (see materials and methods). Presented values were returned to the original scale before transformation. § Assuming round patch shape. ¶ I value data based on 14 random patches. # t test was preformed assuming unequal variances. 1366 CROP SCIENCE, VOL. 46, MAY–JUNE 2006

Fig. 4. The distribution of the different microhabitat types among populated sampled units. Numbers within pie slices indicate the total populated sample units of each category.

exposed, or rocky–stony. Heterogeneity of habitats was C. judaicum was particularly low on the northern slope observed in all tested slopes with the highest SDI value of N. Mea’ra (Table 5). detected at the north facing slope of N. Mea’ra and The frequency of C. judaicum in all habitat types with atthenorthwestslopeofN.A’naba(Table5).TheSDI stony or rocky components was higher than 20% and value in those two microsites are significantly different above 40% in rocky–stony habitats (Fig. 4). The dis- from the lowest SDI value found at the south facing slope tribution of the different habitat types within the two of N. Mea’ra. No significant differences in micro habitat groups of populated and unpopulated sampling units diversity were found between the four slopes of N. was nonrandom as evident from the Pearson test (x2 5 A’naba site (Table 5). The three slopes of N. A’naba with 77, P(x2) , .0001), a hint for differential habitat pref- the intermediate SDI values (eastern, south eastern and erence of the species. Of all populated sampling units, western) had the highest frequency, number of patches, 87.6% were characterized as stony or rocky habitats. and density of C. judaicum in a patch. Frequency of Indeed, out of 191 populated sampling units, 91 were Reproduced from Crop Science. Published by Crop Science Society of America. All copyrights reserved.

Fig. 5. Values of mean 6 S.E. plant density at N. A’naba and N. Mea’ra sites on three sampling dates. The influence of the sampling date factor on plant density was significant P(F) , 0.0001. Tukey Kramer test was applied for comparing plant density between sampling dates at each site. Columns labeled with different letters differ significantly (within sites) at a 5 0.05. BEN-DAVID ET AL.: ECOLOGY OF WILD CICER JUDAICUM 1367

Table 5. Microhabitat spatial pattern and landscape structure of slopes inhabited by C. judaicum at the two tested sites. Slopes means 6 SE of Shannon diversity index together with frequency and patch number per transect are indicated. The classification of microhabitats was made on 475 of 1 m2 units sampled during the 2003 season at N. A’naba and N. Mea’ra. Calculation and analyses were made along each transect separately, each including 25 sampling units. To compare means, Tukey Kramer test was performed. In each column, means labeled with different letters differ significantly at a 5 0.05. Site Slope aspect Shannon diversity index (SDI)†,‡ Frequency‡,§ Mean patch numbers per transect Density within patch N. Mea’ra southern 1.12 6 0.08b 0.22ab 2.88 6 1.72 2.18 6 0.59 northern 1.7 6 0.17a 0.04b¶ 1.50 6 0.70¶ 2.32 6 0.76 N. A’naba eastern 1.46 6 0.10ab 0.35ab 3.25 6 0.95 5.18 6 1.20 south eastern 1.33 6 0.17ab 0.47a 3.50 6 0.50 5.39 6 0.98 western 1.18 6 0.17ab 0.49a 3.50 6 0.50 4.61 6 0.90 north western 1.67 6 0.17a 0.13ab 2.50 6 0.29 3.97 6 1.07 † Index was calculated on eight arbitrary microhabitats (see materials and methods). ‡ Tukey Kramer test was applied for comparing SDI and frequency means between slopes. § Frequency values were transformed before analysis (see materials and methods). Values of means were returned to the original scale before transformation. ¶ Mean is based on four transects.

defined as rocky–stony, 47 as stony, and 28 as stony– cases, Givat Ada and Talmon, populations were found in exposed (Fig. 4). newly constructed sites. It remains to be seen if in the long run the species will survive in the two latter lo- DISCUSSION calities. In traditional farming sites, still prevalent in the Palestinian Authority, C. judaicum plants were observed Cicer judaicum Distribution Range: Size, in olive groves margins (Ebtan and Ras-Karkar). Inter- Boundaries and Internal Structure estingly, C. pinnatifidum, which thrives in similar con- Ecologists concerned with the size of geographic texts, were observed (by Can and Abbo) in Gaziantep range of species pointed out that, in certain cases, province (near Burç) and along the GolbaYi-Adiyaman the range decreases with decreasing elevation (Brown road in eastern Turkey. et al., 1996). Apparently, this observation also holds true About 39% of the survey sites were within nature when comparing the distribution range of C. judaicum reserves. Yet, C. judaicum populations were found only to the present known distribution of other annual Cicer in three out of these 17 protected sites, suggesting poor species (Berger et al., 2003). Accordingly, C. judaicum prospects for in situ conservation of this species. It is distribution area is smaller than the ranges of C. probably safe to say that the high rate of infrastructure reticulatum and C. echinospermum P. H. Davis and con- development in Israel, a densely settled country, puts siderably smaller than those of its two closely related most Israeli populations of the species at risk of extinc- species C. pinnatifidum Jaub. & Spach and C. bijugum (in tion. For instance, such extinction has already occurred order of magnitude difference). around Jerusalem where specimens were collected from The internal structure of the C. judaicum distribution what is now within the city area (e.g., latest collections range explains the irregularity of the population spa- from the 1970s within the Giva’at-Ram campus of the tial scattering. (Fig. 1, 2). Many areas within the overall Hebrew University). Israeli distribution area are uninhabited (zero local Another aspect of natural habitats disturbance is the abundance). This internal structure characteristic, dis- process of fragmentation, an outcome of breaking the cussed in a general theoretical context by Brown et al. continuity of a natural habitat area by zones of con- (1996), makes any attempt to estimate the general fre- struction or other human interference. This includes quency of C. judaicum problematic. Nevertheless, the two components: first, the reduction in habitat area, relatively low abundance of the species is an established and second, the formation of isolated populated islands fact throughout Israel. derived from an original continuous populated space. The absence of C. judaicum from the upper Galilee One possible outcome of such isolation might be limited and the Golan Heights may account for the ecogeo- (or absence of) gene flow, followed by increased dif- graphic adaptive preferences of the species. The appar- ferentiation and/or decreased fitness. In the long run, ent distribution range may reflect avoidance of sites with this process could proceed to full extinction (Frankham relatively low winter temperature, mild to high annual et al., 2002). In the case of C. judaicum, a self-pollinating rainfall, and intermediate altitude (.800 m above sea species with limited cross pollination, severe reduction level in this survey), as expressed by the relatively low in population size might have profound and radical Reproduced from Crop Science. Published by Crop Scienceoccupancy Society of America. All copyrights reserved. of sites scores in the positive quarter of the effects on the population allele pool. In fact, the rela- PCA chart (Fig. 3). tively small distribution range of the species in Israel Cicer judaicum populations were found in different suggests a pessimistic prospect. This subject, how- types of habitats, from Mediterranean forest in Hasole- ever, needs to be considered in a much wider context lim and rocky slopes at N. Mea’ra to planted pine forest of the Mediterranean basin as a whole. This region suf- in Kesalon and Mevo Beitar. We never encountered fers from 95% loss of primary habitats resulting in C. judaicum in a ruderal context such as road sides, the second largest portion of endangered plant spe- settlements, and field edges. Only in one case (Nataf) cies after the tropical rainforests (Myers et al., 2000). plants were found along the edge of a countryside paved The latter authors defined the Mediterranean basin road (seldom sprayed against weeds). In two other as a “hotspot” from a biodiversity standpoint, hosting 1368 CROP SCIENCE, VOL. 46, MAY–JUNE 2006

13 000 endemic species, 50% of the region’s total num- species with lower local density. Across the Israeli sites, ber of plants. planted forests and natural habitats with the “search The distribution map (Fig. 1) highlights the confine- image” formed during this survey were not rare, thereby ment of Israeli C. judaicum populations into a relatively supporting the possibility of such positive correla- narrow range. The world collection, however, includes tion. It seems that C. judaicum may be an interesting accessions from Lebanon and Syria, areas which pres- case to test the above hypothesis due to its apparent ently are beyond our reach. Notably, both C. judaicum low density (in most sites) and relatively restricted geo- and C. pinnatifidum (a closely related annual species) graphic range. are reported to grow sympatrically in the Damascus basin, Syria (van der Maesen, 1972). Cicer judaicum was Population Demography also reported in eastern Turkey (Fig. 1 of Berger et al., The N. Anaba and N. Meara sites (depicted in Fig. 1) 2003). However, in a recent survey in eastern Turkey (by were chosen for a detailed two-season demographic C. Can and S. Abbo), these claims could not be con- assay at the population level. Estimate of local pop- firmed, as not a single C. judaicum population was found ulation density is often being used as a predictor for in Gaziantep, Adiyaman, Urfa, Marash, Diyarbakir, and the overall geographic range of species. The spatial dis- Mardin provinces. Rather, thriving populations of C. tribution of individual C. judaicum plants within the pinnatifidum were easily located in all the above Turkish population space is patchy (Table 5). A major factor provinces. In our view, the pattern observed in eastern determining plant spatial distribution is the mode of Turkey and the complete absence of C. judaicum pop- seed dispersal (Perevolotsky and Polak, 2001). Cicer ulations from the northern parts of Israel calls for re- judaicum is a short-stature plant most of whose pods fall evaluation of herbaria materials, detailed surveys in to the ground and shatter only thereafter, thus minimiz- other eastern Mediterranean countries, and reconsider- ing long distance and even seed dispersal. This may ation of current distribution data of these two taxa. account for the relatively limited size of C. judaicum patches. A second important factor is the occurrence Prediction of Potential C. judaicum Habitats and spatial distribution of safe-sites for germination. Heterogeneous habitats are usually a mosaic of such C. judaicum Cicer Since the biology of (like most wild safe and unsafe patches for the establishment of par- species) is not well known, any information regarding ticular species (Crawley, 1997). factors limiting its distribution is significant. We imple- Density values in annual plant populations may vary mented a correlative predictive model (Robertson et al., between seasons (Noy-Meir et al., 1991b). Because of 2003), which pointed to three areas with a potential to the short time frame of our study, we dealt only with C. judaicum host habitats of , namely, the Judean Moun- intraseasonal changes. In both sites, plant density was tains and foothills, west Samaria, and lower Galilee higher during the early winter compared with mid win- (Fig. 2). Further exploration in those areas might be ter and early spring, apparently for two reasons (Fig. 5): rewarding. Since we have surveyed those three regions first, the intensive cattle grazing that occurs at N. Anaba C. judaicum quite intensively, we suspect that, if present, in early February (N. Mea’ra site was also exposed to must be relatively rare in these areas. Nonetheless, the sheep grazing), and second, the ease of identifying the unpopulated lower Galilee region which was highlighted four-to-five leaf stage plantlets, which are common in by the prediction model is an attractive area because early winter. At this timing, the stems of the common C. judaicum it harbors many locations in which no was grasses have not yet elongated, leaving the C. judaicum recorded thus far. young plants unmasked by the grass canopy, hence easy The use of ecogeographical data from such surveys to find (Fig. 6). Intraseasonal fluctuations were detected may be somewhat problematic. This is because of inher- also in the frequency variable, but those were not sta- ent extrapolation errors and the scattering of meteoro- tistically significant (data not shown). Notably, the cor- logical stations that do not expose the full relevant relation between the density variable and the frequency microclimate spectrum (Guisan and Zimmerman, 2000). variable, although significant (p 5 0.005), is only 0.48. In addition, our work was restricted by political borders Therefore, frequency values will not automatically cor- (to the north and east of the survey area) with the pos- respond to changes in density, particularly in light of the sible result of exposing only part of the true distribution aggregated spatial distribution pattern of the species. picture. This is not exceptional, as political boundaries limitation is a common feature in many studies of Reproduced from Crop Science. Published by Cropspecies Science Society of America. All copyrights reserved. geographic range (Gaston, 1990). Consequently, Populated Patch Descriptors our prediction model results should be approached The north facing slope of N. Mea’ra and the north- with caution. west facing slope of N. Anaba are characterized by A positive correlation between abundance and dis- the highest compositional habitat diversity values (SDI, tribution range is expected when the reference habitat Table 5). Significant portions of these slopes support (in which abundance was measured) is common through- growth of shrubs associated with Mediterranean ga- out the range over which distribution is recorded (Gaston rigue. Consequently, and probably because of competi- and Lawton, 1990). Gaston (1996) claimed that this pos- tion, C. judaicum plants occur therein at low frequency itive correlation may cause some measurement errors values, small number of patches per transect, and lower of underestimating the distribution and occurrence of density within patches (Table 5). BEN-DAVID ET AL.: ECOLOGY OF WILD CICER JUDAICUM 1369

Fig. 6. Young C. judaicum plant photo was taken on December at N. Mea’ra site (a, left panel). A C. judaicum plant among dense grasses, mainly Hordeum spontaneum photo was taken on February at N. A’naba (b, right panel).

Wild Mediterranean grasses [like barley, Hordeum species (Berger et al., 2003). Therefore, this work rep- spontaneum C. Koch, and wheat, Triticum dicoccoides resents a significant step toward better understanding of (Koern. ex Asch. & Graebner)] covering vast tracts of the biology of wild Cicer in general and its annual spe- land across the eastern Mediterranean, are much more cies in particular (Abbo et al., 2003). To make the above abundant compared with wild legumes (e.g., wild lentil information useful for future improvement of the chick- and pea) as evident from seed yield data of wild pea crop, further research is required to determine the populations (Harlan, 1967; Kislev et al., 2004; Ladi- genetic basis of the macro- and microecological pre- zinsky, 1975, 1987). Surprisingly, when measured at the ferences of this and other wild Cicer species. populated patch level, C. judaicum density values are comparable with those measured for wild emmer wheat, T. dicoccoides Aarons. (0.4–4.8 plants per m2), as re- ACKNOWLEDGMENT corded in Ammiad, Israel, a rich site in the heartland of R. Ben-David is indebted to the Israeli Gene Bank for wild wheat range (Noy-Meir et al., 1991b). stipend support.

Microhabitat Characterization REFERENCES Cicer judaicum is much more abundant in microsites Abbo, S., J. Berger, and N. Turner. 2003. The evolution of cultivated chickpea: A series of three genetic bottlenecks limit diversity and with a stony component (Fig. 4). Whether the effect is constrain crop adaptation. Funct. Plant Biol. 30:1081–1087. exerted through soil moisture, modification of grazing Abbo, S., C. Molina, R. Jungmann, M.A. Grusak, Z. Berkovitch, Ruth pressure, competition with more aggressive species, or Reifen, G. Kahl, P. Winter, and R. Reifen. 2005. QTL governing any combination of the above factors still remains to be carotenoid concentration and weight in seeds of chickpea (Cicer determined. Similarly, in wild emmer wheat, a species arietinum L.). Theor. Appl. Genet. 111:185–195. Austin, M.P. 1985. Continuum concept, ordination methods and niche with strong affinity to rocky habitats too, the major theory. Annu. Rev. Ecol. Syst. 16:39–61. determinants of habitat preference are water availability BioGis. 2002. Israel Biodiversity Information System [Online] mediated through water run-off from rocks into the soil, Available by The Hebrew University of Jerusalem. http://www. and a reduction in direct radiation and wind speed af- biogis.huji.ac.il/biogis/static/en/index1.html?param15legal; verified 10 January 2006. fecting evaporation from the soil (Noy-Meir et al., Berger, J., S. Abbo, and N. Turner. 2003. Ecogeography of annual wild 1991a, 1991b). Presumably, in such habitats, C. judaicum Cicer species: The poor state of the world collection. Crop Sci. may germinate early and be ensured better establish- 43:1076–1090. ment than in more exposed and open microhabitats. In Berger, J., and J. Speijers. 2003. An introduction to modern PC-based that case its microhabitat preference could be viewed as data handling and statistical methods. 2nd ed. p. 88–98. CLIMA, Perth, WA. competitive release. Also, stony and rocky surface pro- Brown, J.H., C. Stevens, and D.M. Kaufman. 1996. The geographic vides relative protection from cattle grazing by limiting range: Size, shape, boundaries, and internal structure. Annu. Rev. Reproduced from Crop Science. Published by Crop Sciencemovement, Society of America. All copyrights reserved. while in exposed microhabitat grazing Ecol. Syst. 27:597–623. pressure is stronger (Noy-Meir et al., 1991a). Crawley, M.J. 1997. Plant-herbivore dynamics. p. 401–474. In M.J. Crawley (ed.) Plant ecology. 2nd ed. Blackwell Science Ltd, Oxford, UK. Concluding Remarks Delacy, I.H., K.E. Basford, M. Cooper, J.K. Bull, and C.G. McLaren. 1996. Analysis of multi-environment trials—An historical perspec- This work combined macroecogeographical and mi- tive. p. 39–124. In M. Cooper and G.L. Hammer (ed.) Plant adap- croecological observations aiming to analyze the factors tation and crop improvement. CAB International, Wallingford, UK. that determine the boundaries of C. judaicum distribu- Engels, J.M.M., R.K. Arora, and L. Guarino. 1995. An introduction to tion and its physical population structure in Israel. No plant germplasm exploration and collecting: Planning, methods and such work has ever been reported for any wild Cicer procedures, follow-up. p. 31–63. In L. Guarino et al. (ed.) Collecting 1370 CROP SCIENCE, VOL. 46, MAY–JUNE 2006

plant genetic diversity—Technical guidelines. CAB International, and cultivation. Wageningen Agricultural University, Wageningen, Wallingford, UK. the Netherlands. Environmental Systems Research Institute (ESRI). 2002. Arc- McGarigal, L., and B.J. Marks. 1994. FRAGSTATS Manual: Spatial View. Released 3.2ESRI (http://www.esri.com/index.html; verified pattern analysis program for quantifying landscape structure, 15 December 2005). URL:ftp://ftp.fsl.orst.edu/pub/fragstats.2.0; verified 10 January 2006. Evans, L. 1993. Crop evolution, adaptation and yield. Cambridge Miller, P.R., B.G. McConkey, G.W. Clayton, S.A. Brandt, J.A. Staricka, Univ. Press. Cambridge, UK. A.M. Johnston, G.P. Lafond, B.G. Schatz, D.D. Baltensperger, and FAO. 2006. FAOSTAT [Online] Available by food and agriculture K.E. Neill. 2002. Pulse crop adaptation in the northern Great organization (http://faostat.fao.org/; verified 17 January 2006). Plains. Agron. J. 94:261–272. Frankham, R., J.D. Ballou, and D. Briscoe. 2002. Introduction to Myers, N., A.R. Mittermeier, G.C. Mittermeier, B.A. da Fonseca, and conservation genetics. Cambridge Univ. Press, Cambridge, UK. J. Kent. 2000. Biodiversity hotspots for conservation priorities. Gaston, K.J. 1990. Patterns in the geographical ranges of species. Biol. Nature (London) 403:853–858. Rev. 65:105–129. Noy-Meir, I., M. Agami, E. Cohen, and Y. Anikster. 1991a. Floristic Gaston, K.J., and J.H. Lawton. 1990. Effects of scale and habitat on the and ecological differentiation of habitats within a wild wheat relationship between regional distribution and local abundance. population at Ammiad. Isr. J. Bot. 40:363–384. Oikos 58:329–335. Noy-Meir, I., M. Agami, E. Cohen, and Y. Anikster. 1991b. Changes in Gaston, K.J. 1996. Species-range-size distribution: Patterns, mecha- the population density of wild emmer wheat (Triticum turgidum var. nisms and implications. Trends Ecol. Evol. 11:197–201. dicoccoides) in a Mediterranean grassland. Isr. J. Bot. 40:385–395. Greig-Smith, P. 1964. Quantitative plant ecology. 2nd ed. p. 21–167. Perevolotsky, A., and G. Polak. 2001. Ecology: Theory and the Israeli Butterworth & Co. Ltd., London. Experience, Carta, (Hebrew), Jerusalem. Guisan, A., and N.E. Zimmerman. 2000. Predictive habitat distribu- Ravikovitch, S. 1969. Manual and map of soils of Israel. Magnes Press, tion models in ecology. Ecol. Modell. 135:147–186. The Hebrew University, Jerusalem. Harlan, J.R. 1967. A wild wheat harvest in Turkey. Archaeology 20: Robertson, M.P., C.L. Peter, M.H. Villet, and B.S. Ripley. 2003. Com- 197–201. paring models for predicting species’ potential distributions: A case Kershaw, K.A. 1973. Quantitative and dynamic plant ecology. 2nd ed. study using correlative and mechanistic predictive modeling tech- Edward Arnold Ltd., London. niques. Ecol. Modell. 164:153–167. Kislev, M.E., E. Weiss, and A. Hartmann. 2004. Impetus for sowing Siddique, K.H.M., S.P.Loss, K.L. Regan, and R.L. Jettner. 1999. Adap- and the beginning of agriculture: Ground collecting of wild cereals. tation and seed yield of cool season grain legumes in Mediterra- Proc. Natl. Acad. Sci. USA 101:2692–2695. neanenvironmentsofsouth-westernAustralia.Aust.J.Agric.Res.50: Ladizinsky, G. 1975. Collection of wild cereals in the upper Jordan 375–387. Valley. Econ. Bot. 29:264–267. SAS Institute. 2001. Jmp start statistic. 2rd ed. Duxbury, Pacific Grove, Ladizinsky, G. 1987. Pulse domestication before cultivation. Econ. CA. Bot. 41:60–65. SPSS. 2002. ‘SPSS for Windows.’ (http://www.spss.com/). Ladizinsky, G., and A. Alder. 1976. Genetic relationships among the Sokal, R.R., and F.J. Rohlf. 1997. Biometry: The principles and prac- annual species of Cicer L. Theor. Appl. Genet. 48:197–203. tice of statistics in biological research. 3rd ed. p. 419–422. W.H. Legendre, P. 1993. Spatial autocorrelation: Trouble or new paradigm? Freeman and Company, New York. Ecology 74:1659–1673. Turner, N.C., G.C. Wright, and K.H.M. Siddique. 2001. Adaptation of Lev-Yadun, S., A. Gopher, and S. Abbo. 2000. The cradle of agri- grain legumes (pulses) to water-limited environments. Adv. Agron. culture. Science (Washington, DC) 288:1602–1603. 71:193–231. Magurran, A.E. 1988. Ecological diversity and its measurement. Univ. Ward, J.H., 1963. Hierarchical grouping to optimize an objective Press, Cambridge, UK. function. J. Am. Statist. Assoc. 58:236–244. van der Maesen, L.J.G. 1972. Cicer L., A monograph of the genus, with Zohary, D., and M. Hopf. 2000. Domestication of plants in the Old special reference to the chickpea (Cicer arietinum L.), its ecology World. 3rd ed. Oxford Univ. Press. Oxford, UK. Reproduced from Crop Science. Published by Crop Science Society of America. All copyrights reserved.