Can anthropogenic variables be used as threat proxies for South African plant richness ?

Human demographic and socio-economic measures (anthropogenic variables) reflect the detrimental impact o f humans on plant diversity globally. The Pretoria (PRE) Computerised Information System (PRECIS) of the South African National Biodiversity Institute (SANBI), provided three sets of South Afncan plant richness data, overall (OPR), endemic (EPR), and threatened (TPR), to investigate the relationships between richness and six anthropogenic variables. Spearman’s Rank order correlations, Kruskal Wallis Analysis o f Variance (ANOVA) and Generalized Linear Models (GLZ) were used. Although all three plant richness measures were correlated with anthropogenic variables, individual anthropogenic variables contributed a small fraction to the explained variation in richness. Differences in spatial and temporal scaling of the datasets, or the response to another causal mechanism, may have contributed to this low explained variation. Because more variation was accounted for in OPR than EPR or TPR, OPR is a more suitable surrogate measure of plant biodiversity when investigating the anthropogenic variables used here. Average human density (HD), infrastructure (degree of urbanization and road cover) (LRU) and percentage land area transformed and degraded (LTD) were identified as useful surrogates o f human impacts on OPR. LTD may be a more inclusive human impact measure when conducting analyses of human impacts using OPR. LTD includes the effects of urban expansion, road networks and other land transformation impacts, such as agriculture.


INTRODUCTION
Human actions threaten biological diversity at a global scale.The sources of threat include taxon-specific threats such as exploitation, introduced taxa and various forms of ecosystem degradation, including land transfor mation and pollution (World Resources Institute 2000).Strong evidence indicates correlations between rates of habitat and species disappearance with human demogra phy patterns, such as population density and population growth, and human activities (James 1994;Ceballos & Ehrlich 2002;Harcourt & Parks 2003;Luck et al. 2003).As a result, conservation planners have been urged to integrate biological, socio-economic and human demo graphic data in their assessments to effectively determine real world conservation priorities (Cincotta et al. 2000;Brooks & Thompson 2001).
Human demographic and socio-economic data are often current and easily available (Harcourt & Parks 2003), and may constitute useflil surrogate measures of the proximate threats to certain life forms and taxa.A number of human-related variables are correlated with plant richness measures around the world, both at regional and global scales.These include human demo graphic parameters (e.g.population density, human population growth rate change, poverty and affluence, urbanization), land transformation, land fragmentation and fuel wood consumption (Macdonald 1991;Kerr & Currie 1995;Cincotta et al. 2000;Ceballos & Ehrlich 2002;Liu et al. 2003).
O f these parameters, human population density has been considered a reasonably good indicator of threat of the risk o f species extinction (Thompson & Jones 1999;Harcourt & Parks 2003).Plant population declines are mostly concentrated in areas with either high human densities or high human impact, such as agriculture (Burgess et al. 2002;Ceballos & Ehrlich 2002;Araujo 2003).In turn, rapid and continuing population growth, and the associated human impact on the environment, is ever increasing (Cincotta et al. 2000;World Resources Institute 2000;Liu et al. 2003), necessitating urgent investigation into the relationship between plant richness and human population increase.
Sub-Saharan countries have some o f the highest population growth rates in the world (United Nations Development Programme 2001).This, tied with high human population densities, clearly translates into considerable landscape transformation (James 1994;McKinney 2001).Most of Africa's dense human settle ments, intensive agricultural activities and habitat frag mentation are concentrated in areas of high animal and plant endemism (Balmford et al. 2001;Harcourt et al. 2001;Burgess et al. 2002;Chown et al. 2003).
A strong association between population growth and environmental degradation exists that is mediated, in part, through income.For example, it has been dem onstrated that as human populations grow, agricultural productivity declines, and this in turn raises rural pov erty (Ukpolo 1994).Poverty stricken people are forced to rely heavily on surrounding resources for survival, placing increased pressure on vegetation and plant and animal species in the region (Lucas & Synge 1981;James 1994).Struggling populations in rural areas may then move to the cities (Ukpolo 1994;World Resources Institute 2000) resulting in, for example, urban sprawl that generally involves complete transformation of rela tively large areas (Macdonald 1991;Cincotta et al. 2000;Liu et al. 2003).
On the other end o f the income scale, higher per cap ita income may also lead to environmental degradation (Naidoo & Adamowicz 2001).Areas with a high Gross National Product (GNP) correspond to areas with a high proportion of threatened plants in various countries (Kerr & Currie 1995;Naidoo & Adamowicz 2001).High GNP usually fuels excessive land conversion and resource exploitation, increasing the number of threatened plant and vertebrate taxa (Naidoo & Adamowicz 2001).In many cases, this association between humans and envi ronmental degradation, and humans and species threat, is mutually reinforced by additional anthropogenic mea sures such as household dynamics, urbanization, technol ogy and political instability (World Resources Institute 2000; Liu et al. 2003).In this sense, the combined forces o f human population pressure apply tremendous stress to ecological systems in Africa and other underdeveloped areas (Ukpolo 1994).
Environmental degradation extends beyond the effects of human population growth.Land use and land-cover changes are important elements of the larger problem of global environmental change.Land use impact and the loss of species can often be directly related to the percentage area under urbanization, transformation or fragmentation (Wood et al. 1994;Pfab & Victor 2002).Theobald (2003) found that areas with more than 15% infrastructure development coverage could be deemed highly fragmented and thus impacting negatively on bio diversity, resulting in species loss (Santos et al. 2002;Tschamtke et al. 2002).Correspondingly, roads have been claimed to have a disproportionate effect on biotic diversity (Macdonald 1991;Reyers 2004).Reyers et al. (2001) illustrated that road construction and mainte nance significantly altered surrounding natural habitat and landscapes.In addition, other forms of landscape transformation and degradation such as areas utilized for agriculture and plantations may lower species richness.Although the effect o f high human population density on species richness may be observable through increased road infrastructure and urban expansion, agriculturally degraded or transformed landscapes are likely to occur in lower human density areas, confounding the relation ship between land transformation and human population density.As a result, land-cover changes may also impact species richness patterns.
Identifying and managing threatening processes, threatened areas and taxa at risk of extinction at national scales, in essence, requires interpreting the impacts of human activities on biodiversity at smaller spatial extents (Brooks & Thompson 2001;Gardenfors et al. 2001;Chown et al. 2003;Liu et al. 2003).Incorporating human demographic and socio-economic variables into conservation priority setting procedures will allow more informed conservation decisions to be made (Hannah et al. 1994;Sisk et al. 1994).An understanding o f the rela tionships between these variables and plant richness at national scales is essential to determining the suitability o f such data as surrogate measures to assist in under standing the proximate and ultimate threats to plant taxa (Kerr & Currie 1995;Brooks & Thompson 2001).However, ascertaining the relationships between biodi versity and human impacts are not as straightforward as originally believed.The effects of, and threat posed by, human demographic and socio-economic activities on taxon richness at national scale remain unclear because of the complexity and multiplicity of human activities, as mentioned above.
The impacts o f human activities on biodiversity may also vary between richness measures (Van Rensburg et al. 2002).For example, human impacts may be greater, and therefore more apparent, for threatened and endemic plant taxa than for total species richness.Higher taxon richness has been successfully used as a biodiversity sur rogate in prioritizing areas of conservation importance at both national and continental scales (Fjeldsa 1997;Balmford et al. 2001;Harcourt & Parks 2003;Rouget et al. 2004).But conservation strategies based solely on overall plant richness are often of limited use (Jetz & Rahbek 2002).The large proportion of endemic taxa in southern Africa has been attributed to the diverse ecolog ical conditions, as well as the product of high speciation within a large number of endemic genera (Cowling & Hilton-Taylor 1994).It is essential to include an analysis of threatened and range restricted taxa when classifying areas of conservation priority (Rouget et al. 2004).
In South Africa, the Western Cape has been favoured by human settlement for the past 350 years, leading to substantial land transformation through agricultural and urban development, and alien plant encroachment (Deacon 1992;Rebelo 1992a;Richardson et al. 1996).Subsequently, much of the remainder of southern Africa has also undergone extensive land transformation over the past 100 years (Macdonald 1991;Van Rensburg et al. 2004).People have been attracted to the interior where mineral and fossil fuel resources abound (Deacon 1992).This has resuhed in a multi-faceted combination of human demands and inherent threats across the south ern African landscape (Reyers et al. 2001;Neke & Du Plessis 2004).

Previous work investigating human impacts in South
Africa has focused on the relationships of taxon (or spe cies) richness and human variables for South Africa's birds and, recently, frogs (Van Rensburg et al. 2002;Chown et al. 2003;Van Rensburg et al. 2004;Evans et al. 2006).Their results show that avian species richness and human density were positively correlated, apparently as both responded positively to increasing levels of pri mary productivity (Van Rensburg et al. 2002;Chown et al. 2003;Van Rensburg et al. 2004).Here, we investigate the relationships between three plant richness measures that include all endemic and threatened South African plant taxa and six human demographic and socio-eco nomic variables, to better understand potential and real istic anthropogenic threats to South African plants.

METHODS
Plant distributions of all South African taxa were extracted from the Pretoria National Herbarium (PRE) Computerised Information System (PRECIS) for 1 936 quarter degree squares (QDS) of South Africa.Richness maps were collated for three sets of plant groups.First, overall plant richness (OPR) was collated for South Africa at QDS level.Secondly, a list o f 'endemic' taxa believed to occur only in South Africa, were extracted from the PRECIS database as well as Germishuizen et al. (2006) to produce a richness map for South African endemic taxa.Finally, a measure of threat ened plant richness per QDS was calculated based on a list of threatened taxa, which was extracted from the Threatened Species Programme's database of December 2003 (Threatened Species Programme unpublished data).This list of threatened species as used here (pre liminary 2003 Red List of South African plants) is cur rently in the process of being re-assessed according to the lUCN (2001) Red List Categories & Criteria, by the Threatened Species Programme of the South African National Biodiversity Institute (SANBI).On release, this SANBI Red List will supersede the current 2003 list used.All taxa listed either by Hilton-Taylor (1996) as Endangered (E), or Vulnerable (V) or by lUCN (2001) categories Critically Endangered (CR), Endangered (EN) or Vulnerable (VU), were classified as Threatened.
Taxon richness counts across South Africa's QDSs were square root-transformed following Rebelo & Tansley (1993).Endemic and threatened taxon richness per QDS were also square root-transformed and cor rected for total plant richness by dividing by total plant richness in that QDS (following Rebelo & Tansley 1993) yielding standardized endemic plant richness (EPR) and standardized threatened plant richness (TPR) at a QDS scale across South Afnca.

South African anthropogenic data
Six human and socio-economic variables (anthro pogenic variables from here on) were extracted from various data sources at a national scale in South Africa.These include; 1, human population density, 2; human population growlh rate change; 3, a poverty index; 4, an affluence measure; 5, infrastructure; and 6, the degree of land transformation and degradation.Human popu lation density, human population change, poverty, and affluence data were derived from magisterial district data (Central Statistical Service 1995, 1998), whereas all land-cover and transformation data were collated from the National Land-Cover (NLC) database (Fairbanks & Thompson 1996;Fairbanks et al. 2000).To standardize the scale o f this data with the plant distributional data, data were converted to a spatial scale at the QDS level (25 X 25 km') using ESRI Arc View GIS 3.2.The 1996 South African population census data (Central Statistical Service 1998) were used to esti mate the weighted average population density per QDS (human density-HD).Human density was denoted as the average number of people/km* within each QDS.The average percentage increase or decrease of human population per QDS (human growth rate change-HC) over the period 1996to 2001(Central Statistical Service 1998;Rouget et al. 2004) was used as a direct proxy for the impact of human population growth on the environ ment.
A poverty index (economic poverty-EP) was esti mated as the proportion of people per municipality earn ing less than R200 per month (Central Statistical Service 1998).The United Nations Development Programme South Africa (2003) report indicated that people earning less than R354 per month could be regarded as earning below the poverty line.The census data uses broad cat egories o f which 'less than R200/month' together with the 'no income' category are regarded as earning below the poverty level.This allowed the computation of a weighted average of the proportion of people per QDS earning less than R200/month.
A measure of economic affluence (EA) defined as the weighted average Gross Geographic Product (GGP) per capita income per QDS, was based on GGP obtained for all South African magisterial districts (Central Statistical Service 1995).GGP represents 'the remuneration received by the production factors-land, labour capital and entrepreneurship for their participation in produc tion within a defined area' (Central Statistical Service 1995).The Central Statistical Service (1995) provides 1994 estimates of GGP and remuneration of employees by magisterial district in South African Rand (R).Finescale spatial Gross National Product (GNP) data for South Africa were not available, forcing the use of GGP data, which represents the finest-scale data available for South Africa and was incorporated in the current analy sis rather than GNP data used by other authors (Kerr & Currie 1995;Naidoo & Adamowicz 2001).
To obtain a measure to investigate the effects of urbanization and related development (e.g.industry) on plant richness in South Africa, infrastructure coverage (in the form of the percentage of a QDS covered by road and urbanized areas) was estimated (land-cover, roads and urban, LRU).Infrastructure data were extracted from the NLC database (Fairbanks & Thompson 1996).The extent of the urban area was extracted from all types o f 'urban^uilt-up land' land-cover type (= land-cover type 24-30; Fairbanks & Thompson 1996) in the NLC database.A buffered road network for South Africa was obtained from Reyers et al. (2001) representing various buffered road types in South Africa.
The extent of land transformation (land-cover, trans formed and degraded (LTD) was obtained from the NLC database, by calculating and summing the percentage o f each land-cover class in each QDS, based on the six transformed land-cover classes identified by Fairbanks & Thompson (1996) and Fairbanks et al. (2000).These classes were based on seasonally standardized Landsat TM satellite imagery captured primarily during 1994-1995 and included anthropogenic effects such as forest plantations, artificial water bodies, urban/built-up areas, cultivated lands, degraded land as well as mines/quarries.LTD may be a more inclusive human impact measure than HD or LRU alone (when available), as it includes the effects of human population density on urban expan sion, road networks and the effects of other forms of land transformation that occur in lower human density areas but that may significantly affect species richness (e.g.agricultural, industrial and other land transforma tions, such as plantations).
The values of the weighted average anthropogenic variables calculated for each QDS across South Africa (Table 1) conform reasonably well to human statistics presented by the Development Bank of South Africa (DBSA) (2000).This suggests that the rescaled (to weighted average values for each QDS across South Africa) human demographic variables used here do not noticeably differ from the provincial scale data issued by DBSA.

Statistical analysis
All three measures o f richness (OPR, EPR and TPR) and the six anthropogenic variables (HD, HC, EP, EA, LRU, LTD) were log-transformed for statistical analyses.Kruskal Wallis Analysis of Variance (ANOVA) by Ranks and Spearman's R Rank order correlations (Zar 1996) were used to test for statistical differences and correla tions, respectively, between measures.
Generalized Linear Models (GLZ; McCullagh & Nelder 1989) were used to assess the relationship between OPR and EPR, and between OPR and TPR, as well as between each of the three measures of plant richness and each anthropogenic variable independently.Anthropogenic variables were not included simulta neously in the analysis because of high correlations between some variables (which may lead to collinearity in the model), the substantial differences in the manner in which the variables were measured and subsequently rescaled, and because we wanted to examine the con tribution of each variable to explained deviance in each richness variable independently, to assess its poten tial as a surrogate measure of anthropogenic impact.Because the measures of plant richness were in the form of counts, a Poisson distribution with a logarithmic link function was used in the GLZs (Maggini et al. 2002).A goodness-of-fit test (a deviance statistic), which yields the proportion of deviance explained (similar to an R^ value) by the variable in the GLZ (McCullagh & Nelder 1989) was used to determine which anthropogenic vari able contributed the most to explained deviance in the richness variable used.All statistical analyses were based on analytical subroutines in STATISTICA version 6.1 (StatSoft 2001).

RESULTS
Based on the 2003 data extracted from the PRECIS data set, ± 11 200 of the 22 000 recognized South African plant taxa are endemic to South Africa.The preliminary 2003 Red List regarded about 1 900 taxa as threatened with extinction (Threatened).The Western Cape and Eastern Cape proved to have the highest num ber of plant taxa, endemic as well as threatened (Table 1; Figure 1), mainly in the Cape Floristic Region.
Considering the anthropogenic measures, Gauteng displayed the highest human variables except for eco nomic poverty (EP) and economic affluence (EA-GGP/capita).KwaZulu-Natal, was highlighted as the province with the highest proportion of poverty (EP), and Northern Cape highlighted with the highest aver age GGP/capita (EA), mainly the result of lower human population density and large GGP contributors (mining and quarrying, data not shown) in the province (Table 1).Excluding 'per capita' from the GGP measure, Gauteng Province was the largest contributor to South Africa's economy.
Human densities (HD) and high human growth rate change (HC) were evident mostly in the large metro politan areas such as City of Cape Town, eThekwini Municipality (city of Durban) and municipalities in Gauteng Province (Figure 2A, B).EP in turn was higher in areas in the North-West, Eastern Cape and northeast ern KwaZulu-Natal.EA (affluence/capita) was high in areas with low human density, e.g.Northern Cape (Figure 2C, D).Gauteng Province yielded QDSs with the highest levels of urbanization and road coverage, with Western Cape, Eastern Cape (Transkei), North-West and Free State yielding areas o f high transformation and/ or degradation (Figure 2E, F).

Relationships between species richness measures
Overall plant richness (OPR), endemic plant richness (EPR) and threatened plant richness (TPR) were all sig nificantly different from one another across South Africa (Kruskal Wallis //j = 4070; P < 0.001).Nonetheless, EPR was Strongly significantly correlated with OPR (Spearman R = 0.807; n = 1936; P < 0.05), with EPR (dependant variable) contributing to 63% (Pearson X* = 3468.52;d.f.= 1935; P < 0.001) of the total deviance in OPR.OPR and TPR were also strongly correlated, but not as strongly as the relationship between OPR and EPR (Spearman R = 0.679; n = 1936; P < 0.05), with the regression model explaining 61% (Pearson X* = 724.76;d.f.= 1935; P < 0.01) of the total deviance.The strong correlation and percentage deviance explained for over all plant richness with both EPR and TPR indicates, not surprisingly, that endemic and threatened South African taxa are relatively strongly dependent on total taxon richness.Despite this strong link between richness mea sures, ± 40% of the variation in overall plant richness remains unaccounted for when including either endemic or threatened taxa.We felt it prudent to investigate the relationships between each plant richness variable and the six anthropogenic variables, given that the richness measures varied significantly from one another across QDSs and because human impacts on endemic and threatened taxa may differ fi-om overall plant richness.

Relationships between human demographic and socio-economic variables
The anthropogenic variables were all strongly signifi cantly different fi-om one another (Kruskal Wallis H = 7307.40;P < 0.001).All variables were positively cor related with one another except for the measure of afflu ence (EA-GGP/capita), which was negatively corre lated with all anthropogenic variables calculated within the current study (Table 2).The poverty index (EP) and human growth rate change (HC) were the only two vari ables that were not significantly correlated with each other.Human density (HD) was strongly positively cor related with all variables, except EA (Table 2), suggest ing that this variable may be the single, most inclusive anthropogenic variable to include in investigations of human impacts on flora.Human density was particularly strongly correlated with both land transformation (LTD) and roads and urbanization (LRU).

The effects o f anthropogenic variables on species richness measures
Statistical analysis indicated varying relationships between OPR and the six anthropogenic variables.Plant richness (OPR) was positively correlated with human density (HD), human population growth rate change (HC), land-cover roads and urban (LRU) and land-cover transformed and degraded (LTD).Economic poverty (EP) and economic aflfluence (EA) were negatively corre lated with plant richness, indicating that the effect of these variables on OPR was inverse to the effect of the remaining anthropogenic variables (Table 3).Although HD and LTD were strongly correlated (Table 2), LTD was able to contribute 2% more than HD to explained deviance in OPR (Table 3).No single anthropogenic variable accounted for more than 10.10% deviance explained in OPR (Table 3).Therefore, the observed effect of individual anthropogenic variables on OPR was small, but significant.Endemic plant richness (EPR) and the anthropogenic variables were all weakly correlated with one another (Table 3).Only the poverty (EP) variable, which was weakly negatively correlated with EPR, explained a significant proportion o f the deviance (10.50%) in EPR (Table 3).Other than EP, the anthropogenic variables used here do not appear to contribute sufficiently to understanding the effects of human impacts on endemic plant species.
Threatened plant richness (TPR) and the human vari ables were all weakly significantly correlated (Table 3).Poverty (EP) and economic affluence (EA) were again negatively correlated with the plant richness variable (TPR).Only these variables contributed significantly to explained deviance in TPR, although the contribution was minimal (Table 3).

DISCUSSION
Strong evidence links rates of habitat and species loss with human demographic patterns (Cincotta et al. 2000;Ceballos & Ehrlich 2002;Luck et al. 2003).Given South Africa's low economic growth during the past decade (2.7%), increased population size (2.2% per annum) and increasing demand on urbanized areas (DBSA 2000;Statistics South Africa 2002;Chown et al. 2003), anthro pogenic effects on species should be readily apparent.None of the anthropogenic variables measured here contributed more than 10% towards explaining varia tion in any o f the three species richness measures used.Nonetheless, the contribution of single, anthropogenic variables on overall species richness was both signifi cant and observable.It is important to remember that anthropogenic variables entered in combination, may contribute substantially more than individual variables.This analysis is, however, beyond the scope of the cur rent paper.
Various studies have found a significant, albeit weak, relationship between human population size, human population growth, poverty, per capita income, urbaniza tion and species per country (Ehrlich & Holden 1971;Kerr & Currie 1995;Cincotta et al. 2000).Balmford et al. (2001) reported a marked congruence of high spe cies richness and human population density across the African continent.Chown et al. (2003) state that a strong significant relationship exists between South African birds and human population density (R^ = 0.67) at a quarter-degree scale that is driven, in part, by available energy (Evans et al. 2006).Moreover, strong evidence supports remarkably strong correlations between plant richness and potential evapotranspiration, annual pre cipitation, as well as geographic variation in plant taxo nomic richness (Rutherford & Westfall 1986;Currie & Paquin 1987;O'Brien et al. 1998O'Brien et al. , 2000;;Van Rensburg et al. 2002).Yet none of the human predictors under consideration here had any such strong relationship with any of the richness variables, even though weak signifi cant correlation was present.
The discrepancy between temporal scales o f the data may have contributed to the Weak relationships that were detected between richness and anthropogenic variables in this study.The time scale o f the plant richness distri bution data range from specimens collected in the early 1700s until the present time.Most collections occurred in and prior to 1970 with only 7% of all the specimen collections occurring since 1990 (PRECIS unpublished data).Furthermore, plant distribution data are rarely rep resentative and accurate, and in most cases old and out of date (Rebelo 1992a, b;Freitag & Van Jaarsveld 1997;Maddock & Du Plessis 1999;Maddock & Samways 2000;Rouget et al. 2004).Conversely, the human and socio-economic variable data used in the current study generally date from 1994 to present.
Other than time scale differences in data collection, another major factor influencing statistical results and analysis could be the different spatial scales of the biotic and anthropogenic databases.Most o f the anthropogenic data were measured at magisterial district scale and con sequently were transformed by up-or down-scaling to QDS scale and converted to weighted averages.A QDS is often too large to reflect finer-scale topographical and vegetation differences and will most likely not reflect many of the finer interactions between human predic tors and plant richness measures (Rebelo & Tansley 1993;Van Rensburg et al. 2004).However, this effect is unlikely to be substantial as the summarized statis tics for each province, using the data rescaled to QDS (see Table 1), are remarkably similar to the results of the Development Bank o f South Africa issued at provincial scale (DBSA 2000).Notwithstanding these temporal and spatial scaling issues, all the databases are still useful to investigate the presence o f broader spatial scale relation ships and trends, albeit if finer scale interactions may be weakened.
Finally, other more important causal mechanisms may dominate plant richness patterns, e.g.climatic variables, topographic variables and (d iv e rsity (Bailey et al. 2002).Indeed, Evans et al. (2006) suggest that available energy may be driving the observed patterns in bird spe cies richness and human population change.
Even though conservationists are urged to include human variables when setting conservation priorities, human population variables are imperfect indicators of risk to biodiversity (Thompson & Jones 1999;Cincotta et al. 2000).Indeed, the results of this study show that the strength o f the relationships between human vari ables were weak, albeit significant, and the strength of the relationship did not improve when endemic or threat ened taxa were used in the analyses instead of overall plant species richness (OPR).Thompson & Jones (1999) found that human popula tion density accounted for almost 35% of the variability in rare and threatened plant loss in a study conducted in Britain.Our study shows that human population density did explain a significant proportion of the variation in OPR but that its contribution to explained variation in endemic (EPR) or threatened (TPR) plant species rich ness was not significant.Therefore, although there is a strong link between the three richness measures used here, OPR appears to be the most suited richness surro gate when using the anthropogenic variables included in this study.Nonetheless, it remains important to consider how human impacts may affect endemic and threatened plant richness patterns.This study shows that the only suitable anthropogenic variable to use for EPR and TPR is economic poverty.This was the only anthropogenic measure to show a significant relationship with these richness measures.In addition, TPR data used here are based on preliminary 2003 data, which will be super seded by the Red List due in 2007 (Threatened Species Programme, SANBI).This will allow for more up-todate richness data, as well as updated threat data and subsequent threat analysis.The results from the present study should be verified when the updated SANBI Red List becomes available.
Although clear evidence for different measures of human impact affecting plant extinction have been highlighted, it is diflficuh to assign a risk value to the impacts of these measures (Kerr & Currie 1995;Czech & Krausman 1997;Chertow 2001;Ceballos & Ehrlich 2002).Anthropogenic activities form a complex web of threats that is influenced by various socio-economic and political factors, e.g.national policies, economic condi tions and a host of other factors varying among nations (Macdonald 1991;Kerr «& Currie 1995;DBSA 2000;McKinney 2001;O 'Neill et al. 2001;Liu et al. 2003).Testing certain human threat predictors (as used in the current study) should not be taken to mean that other anthropogenic variables are insignificant, as many addi tional human predictors are also extremely important at local scales (Macdonald 1991;Rouget et al. 2004).
For the current study, human population density (HD), land-cover that is transformed and degraded (LTD) and land-cover roads and urban areas (LRU), appear to be useful measures of anthropogenic impacts on OPR.The pooled strength of the relationship between these vari ables and OPR may not increase as HD is strongly cor related with LTD and LRU.LTD is, however, a more inclusive human impact measure than HD or LRU when conducting analyses of human impacts using OPR.The analyses show that slightly more variation in OPR is explained by this variable than for HD.Also, the effects of large human populations on urban expansion and road networks are included in HD and LRU.However, other forms o f land transformation that occur in lower human density areas that may significantly affect species rich ness, such as, areas under agricultural or industrial use or plantations, are not included in HD and LRU.These impacts are included in LTD.Plant population decreases are concentrated in areas with high human densities or high human impact, such as agricultural areas (Burgess et al. 2002;Ceballos «fe Ehrlich 2002;Araujo 2003), also suggesting that the use of LTD may yield more informa tion on human impacts than HD and LRU alone.It has been shown that human density is clearly a proximate threat, whereas agriculture, urbanization, land trans formation, and roads are ultimate threats (Thompson & Jones 1999).Further analysis is required to ascertain whether other human measures are proximate or ultimate threats.
The differing anthropogenic threats should be assessed individually for each province prior to any risk assessment.Plant richness and endemism are not evenly distributed across South Africa, and most o f these endemic taxa are confined to the predominantly winter rainfall Fynbos and Succulent Karoo Biomes (Cowling & Hilton-Taylor 1994;Figure lA, B).Also, as shown here, anthropogenic variables vary substantially between provinces.For example, Gauteng is an important eco nomic region in terms of business and industrial devel opment, mining and agriculture, and is undergoing rapid expansion of urban areas, constituting the most serious threat to plant populations in this province (Phab & Victor 2002).
It is self-evident of the conservation movement today, that conservation targets are set by incorporating a wide variety o f suitable data, ranging from taxa/species infor mation, land types and habitat types (Pressey et al. 2003;Rouget et al. 2004).Using anthropogenic variables to improve conservation priority, setting procedures for plants is somewhat confounded by the low correlation between anthropogenic and richness variables found in the current study.Studies that do use anthropogenic vari ables as threat predictors for taxa should not be taken up haphazardly (Cincotta et al. 2000).To further assist in identifying South African plant taxa threatened by human activity, additional research is clearly required.Analysis of different human and socio-economic vari ables may well yield substantially different results.Undertaking a more detailed study on the effects o f cli mate, water energy dynamics, and topography on vari ous species richness measures (not only woody tree and shrub richness; see O 'Brien et al. 1998), could perhaps provide insight into processes driving species distribu tion, and subsequently assist in identifying threats to these species.Furthermore, it would perhaps be more relevant to investigate human threats and its impact on broader units (e.g.vegetation type, habitat units) for which the QDS data is adequately suited.The analysis performed here provides insight into the relationships (or lack thereof) between human and socio-economic vari ables as used in the current study, and provides conser vation planners with a better understanding of potential anthropogenic variables that could be regarded as threat proxies for plant taxa.
area: approximate no.South African endemic taxa occurring in particular region.Endemic taxa to area: approximation of taxa endemic to area, reported to occur only in area.Threatened and endemic taxa: no.threatened taxa [Critically Endangered (CR), Endangered (EN), Vulnerable (VU) (lUCN 2001), Endangered (E), and Vulnerable (V) pre-1994 criteria] based on preliminary 2003 Red List (Threatened Species Programme 2003 unpubl.data).Information on average anthropogenic variables for each o f areas calculated for all provincial quarter-degree squares (QDS), with abbreviations as follows: HD, human density (people/km-); HC, human growth rate change (% increase/decrease o f people); EP, economic poverty (proportion of people earning < R200/month); EA, economic affluence [Rands (GGP) per capita x 10*]; LRU, land use roads and urbanization (% area under urban or road); LTD, land-cover transformed and degraded (% area transformed or degraded).

FIGURE 1 .
FIGURE 1.-Patterns o f plant richness across South Africa, based on distribution data from Pretoria National Herbarium Computerised Information System (PRECIS) presented at quarter-degree square (QDS, 25 X 25 km).A, overall plant richness (OPR); B, endemic plant richness (EPR); C, threatened plant richness (TPR).Darker shades represent higher plant richness for each QDS.Threatened plant richness was based on the preliminary 2003 Red List of South African plants.Note: this map should not be used for pro vincial conservation purposes until our results are verified by the updated Red List (shortly to be released by SANBI).

FIGURE 2 .
FIGURE 2.-Spatial representation per QDS across South Africa.A, human population density (HD); B, human growth rate change (HC); C, economic poverty (EP); D, measure o f economic affluence (EA) denoted as the weighted average GGP/capita (EA); E, infrastructure of landcover road and urban areas (LRU); and F, degree o f land-cover transformed and degraded (LTD).Darker shades represent higher impacts.