Ecological Archives E094-222-A1

Anne Blach-Overgaard, W. Daniel Kissling, John Dransfield, Henrik Balslev, Jens-Christian Svenning. 2013. Multimillion-year climatic effects on palm species diversity in Africa. Ecology 94:2426–2435. http://dx.doi.org/10.1890/12-1577.1

Appendix A. Detailed description of methods for obtaining palm species distribution and richness patterns.

A.1 Database of palm species occurrences

Occurrence records for 62 African palm species (95% of total species recognized for Africa; Govaerts et al. 2013) were extracted from a comprehensive geographical database compiled from various sources: (1) herbarium collections (the Royal Botanic Gardens, Kew (K); the National Botanic Garden of Belgium (BR); the National Herbarium of the Netherlands (WAG); the Missouri Botanical Garden (MO); the National Herbarium of Namibia (WIND); and herbarium collections accessed through the GBIF data portal such as Nationaal Herbarium Nederland (NHN); Botanic Garden and Botanical Museum Berlin-Dahlem (BGBM), European Environment Agency (EEA), Fairchild Tropical Botanic Garden (FTG), Real Jardin Botanico Madrid (MA), International Botanical Collections GBIF-Sweden (S), Herbarium Hamburgense (HBG), Herbarium of the University of Aarhus (AAU), Museum National d'Histoire Naturelle (MNHN), National Herbarium of New South Wales (NSW), Royal Museum of Central Africa (RMCA); (2) literature surveys e.g., palm revisions, African floras, and other publications on African biotas (Corkill 1949, Russell 1968, Obermeyer and Strey 1969, Wicht 1969, Moll 1972, Otedoh 1975, Markham and Babbedge 1979, Bullock 1980, Otedoh 1982, Sneed 1983, Johnson 1984, Profizi 1985, Barbier 1985, Sugiyama 1985, Dransfield 1986, Cunningham and Milton 1987, Okolo 1988, Beentje 1990, Johnson 1991, Brown 1991, Kinnaird 1992, Medley 1993, Konstant et al. 1995, Tuley 1995, Gibbons and Spanner 1996, Lambert 1998, Wahungu 1998, Yamakoshi 1998, Newton-Fisher 1999, Barot and Gignoux 1999, Hovestadt et al. 1999, Schulenberg et al. 1999, Mollet et al. 2000, Strohbach 2000, Bornkamm et al. 2000, Sambou et al. 2002, Stave et al. 2003, van Gemerden et al. 2003, Bayton et al. 2003, Sunderland 2003, Ali 2004, Amwatta 2004, Eilu et al. 2004, Baranga 2004, Kakudidi 2004, Ford and Bealy 2004, van Valkenburg and Dransfield 2004, McCullough 2004, Ogunjemite et al. 2005, Alonso et al. 2005, McCullough et al. 2005, Kgathi et al. 2005, Fujioka 2005, Stave et al. 2006, Bayton et al. 2006, Nangendo et al. 2006, Kolongo et al. 2006, Wong et al. 2006, Wright et al. 2006a, Wright et al. 2006b, Hoke et al. 2007, McCullough et al. 2007, Kouassi et al. 2008, Mwaura and Kaburu 2009, Ibrahim and Baker 2009, Kouassi et al. 2009, Shapcott et al. 2009, Bayton and Ouedraogo 2009); (3) private databases and observations held by field botanists (see acknowledgements); and (4) Google Earth satellite imagery used for very conspicuous species such as Elaeis guineensis, Hyphaene petersiana and Phoenix reclinata (cf., Dransfield et al. 2008b, Blach-Overgaard et al. 2009). Additionally, we have accessed the following websites to retrieve information on palm occurrences: www.zimbabweflora.co.zw (last accessed 9 March 2009); www.aluka.org (last accessed December 2007); www.unesco.org (last accessed December 2007); www.iucnredlist.org (last accessed December 2007).

Each individual record was meticulously scrutinized for any geographical or taxonomic issues. This included the correction of faulty georeferences (by ABO) and the rectification of misidentifications (by JD). The full database has 5500 data entries of which 2568 are geographically unique. For the current study, we modeled species distributions at 10 × 10 km resolution leaving 2138 unique records for 62 palm species in total (Table A1). African palms were classified into two groups according to their ecology, namely rainforest and open-habitat palms (Otedoh 1982, Tuley 1995, Sunderland 2007, van Valkenburg et al. 2008, van Valkenburg and Sunderland 2008). The rainforest group consisted strictly of rainforest palm species, species which are exclusively found in tropical rainforest and never in other habitats. The remaining species (except for three, see below) were allocated to the open-habitat group which encompassed palms tolerant of exposed and drier conditions. This includes habitats from dry desert to savanna woodland, but mainly in areas with high water tables or with good soil water retention, or in the riparian zones along waterways (Dransfield 1988, Tuley 1995, Dransfield et al. 2008a). Three species (Elaeis guineensis, Jubaeopsis caffra and Phoenix reclinata) were not included in any ecological group, but still counted towards the total species richness estimate. The two former species can occur in both rainforest and open habitats and J. caffra occurs in evergreen coastal forest which does not fall within any of the two ecological groups (Table A1).

A.2 Predictor variables for modeling individual palm species distributions

We selected a wide set of environmental predictors to model the geographic distributions of individual palm species across the African continent. To account for climatic constraints, we selected all 19 climatic variables from the Worldclim database (Hijmans et al. 2005) and added water balance (computed as the annual mean of monthly differences between precipitation and potential evapotranspiration) based on climate data provided by the CRU CL 2.0 data set (New et al. 2002). All climate layers were originally at 5' resolution. We further used six non-climatic environmental variables: (1) the MODIS vegetation continuous field product (VCF) from 2001 representing percent tree cover at 500-m resolution (Hansen et al. 2002); (2) the annual mean normalized difference vegetation index (NDVI) at 10' resolution composed of monthly NOAA-AVHRR satellite images averaged over an 18 year period (1982–2000 excluding 1994) as a proxy for net primary productivity (EDIT Geoplatform 2010); (3) the annual mean and annual standard deviation of the Microwave Quickscat (Qscat) data available in 3-day composites (further processed to create average monthly composites) from the years 2000–2008 at 2.25 km resolution (Long et al. 2001) which represent surface roughness and moisture and are thus related to vegetation structure (Saatchi et al. 2008); (4) the slope of the digital elevation model derived from the U.S. geological Survey (USGS 2008) describing the maximum change in elevation between each cell and its eight neighbors at 1 km resolution; (5) the GLOBCOVER land cover types derived from the European Space Agency environmental satellite mission at 300 m resolution (ESA and MEDIAS-France 2009), and (6) the soil type (SOIL) extracted from the Harmonized World Soil Database at 30" resolution (FAO/IIASA/ISRIC/ISSCAS/JRC 2008). Finally, we included the recent influence of anthropogenic activities on palm species distributions by using (1) human population density derived from population statistics for the year 2000 at 2.5' resolution (CIESIN and CIAT 2005), and (2) the human influence index, consisting of values ranging for Africa from low (0) to high (64) human influence, estimated per grid cell at 1 × 1 km based on several human impact proxy layers such as population density, land transformations, electric power infrastructure and accessibility from roads, rivers and coastlines (Sanderson et al. 2002). All environmental datasets were projected to the Lambert Azimuthal Equal area projection before further geoprocessing.

To model species distributions, we extracted mean values for all continuous predictor variables at 10 × 10 km resolution using the function Zonal statistics in ArcGIS 9.3 (ESRI, Redlands, CA, USA). For the categorical layers (GLOBCOVER and SOIL) we used the tool Thematic Summary Raster in the ArcGIS extension Hawth's Tool (version 3.26) to estimate the area covered by each category in each grid cell to compute its frequency. From the GLOBCOVER layer, we only used the frequency of the category water bodies to represent the hydrology across the continent as many palms are associated with the riparian zone. From SOIL, we used the frequency of the 10 most frequent soil types (≥ 3% coverage) to represent the edaphic variability. The reason for using soil type (and not individual soil characteristics such as e.g., pH) was to have a full coverage across the African continent. To avoid any potential multicollinearity issues in the later modeling of species distributions we computed a Principal Component Analysis (PCA) using all climatic and non-climatic environmental variables. We extracted the principal components with eigenvalues ≥ 1 (n = 11) accounting for 79.43% of the variation in the data.

In addition to climatic and non-climatic environmental variables represented by the PCA axes, we further included broad-scale spatial constraints that go beyond the measured environmental effects of environmental control. For this, we used spatial filters (Griffith 2003, De Marco et al. 2008) as these have recently been found to improve model performance on African palm distributions (Blach-Overgaard et al. 2010). The spatial filters are eigenvectors derived from a Principal Coordinate Analysis (Borcard and Legendre 2002) based on a pairwise distance matrix computed from the geographical coordinates (latitude and longitude) of the centroids of all grid cells Spatial filters are orthogonal variables which represent the spatial relationship amongst spatial units (grids) at various scales from broad to fine-scale spatial patterns. The spatial filters were computed in SAM 3.0 (Rangel et al. 2006) using default settings and further geoprocessed as described in Blach-Overgaard et al. (2010) and for the current study resampled to 10 × 10 km resolution. We extracted the first 11 spatial filters because they represent broad-to-medium scale spatial constraints and balance the number of principal components represented by the environmental variables (cf., Blach-Overgaard et al. 2010).

A.3 Species distribution modeling

We used species distribution modeling (SDM) to estimate the geographic distributions of individual palm species across Africa. Recent years have shown an increase in the implementation of SDM methods in ecology and conservation (Guisan and Zimmermann 2000, Guisan and Thuiller 2005) and detailed comparative analyses have examined the strength and weaknesses of these techniques (Elith et al. 2006, Elith and Graham 2009). Due to differences in distribution records among palm species (Table A1), we used a three-way approach to estimate species distributions at a continental scale. For species with ≥30 records we used advanced SDM methods, for species with low sample sizes (30 > n ≥ 5) we used a bioclimatic envelope model, and for species with < 5 records we used the observed (rather than modeled) distribution. The details of the SDM modeling approaches are explained below.

For species with ≥30 records (n = 25, Table A1), we used advanced SDM methods and combined them in an ensemble approach to minimize model uncertainties (Araújo and New 2007). We used two regression-based modeling techniques and two machine-learning methods commonly used in SDM (e.g., Elith et al. 2010). These were generalized linear models (GLM), generalized additive models (GAM), and generalized boosting models (GBM) as implemented in the R package ‘Biomod' version 1.1-5 (Thuiller et al. 2009) in R 2.11.1 (R Development Core Team 2010). Additionally, we used MAXENT (Phillips et al. 2006) as it has been found to outperform many other SDM algorithms (Elith et al. 2006, Elith and Graham 2009). The GLM were fitted with quadratic and linear term responses to the predictors using the Akaike information criterion to select models of increasing fit. For GAM, we used a spline function with a smoothing term of 4 (equivalent to a polynomial of 3rd degree). For the GBM, we chose default settings of BIOMOD running maximum 3000 trees and five cross validations to select the optimal number of trees. Finally, for MAXENT we also chose default settings as they have been shown to provide overall robust results (Phillips and Dudik 2008). For all models, we used 10,000 randomly selected pseudo-absences as a random selection of pseudo-absences has been shown to outperform other selection techniques (Wisz and Guisan 2009).

For each species the presence data was randomly divided in calibration (80%) and test (20%) data sets for model evaluation. This procedure was done twice for each algorithm. To assess the performance of the algorithms we used the threshold-independent measure area under the receiver operating characteristic curve (AUC) (Lobo et al. 2008) and the threshold-dependent true skill statistic (TSS) (Allouche et al. 2006). We averaged the AUC and TSS for each model-run using the GAM, GLM, GBM, and MAXENT algorithms and subsequently tested their overall performance using a paired one-tailed Wilcoxon's signed-rank test with species as the sample unit. Models with AUC values ≥ 0.75 are considered to produce good reliable predictions (Elith 2000). TSS ranges from -1 to +1, where +1 indicates a perfect model fit and values ≤ 0 indicates models which are no better than random (Allouche et al. 2006). We used the detailed TSS classifying index as implemented in ‘Biomod': 0.2–0.4 poor; 0.4–0.6 fair; 0.6–0.8 good; and 0.8–1.0 excellent. All palm species distributions modeled by MAXENT, GBM, GAM, and GLM were statistically well-predicted according to AUC (mean AUC ≥ 0.857, except for Hyphaene thebaica (AUCGAM = 0.774, AUCGLM = 0.781)) and TSS (mean TSS ≥ 0.635 except for H. thebaica (TSSGAM = 0.564, TSSGLM = 0.578) with MAXENT performing significantly better than the three remaining models (which according to the Wilcoxon signed-rank test were not significantly different, Table A2). Within the ‘Biomod' environment, it is possible to assess which model has the highest predictive accuracy per run based on AUC or TSS. We used these frequencies to assess if one modeling technique more often than expected by chance outperformed the remaining models testing the model frequencies against the null expectation that the models predict equally well using a Goodness of fit-(G-)test (Sokal and Rohlf 1995). We found that GBM more often than expected by random outperformed GAM and GLM (Table A3). In addition to the statistical model evaluation every single predicted distribution was visually inspected. We found that individual species distributions for some species were highly overpredicted according to our expert knowledge on the African palm distributions, and for other species the distributions exhibited spurious patterns when predicted by GLM or GAM. We found these inconsistencies for other settings also (different spline settings in GAM and different term responses in GLM, result not shown), so we concluded that these inconsistencies were not artifacts from the model tuning. Based on the poor performance (verified visually) of GAM and GLM in estimating species distributions statistically (Table A2, A3), we decided to exclude the distributions predicted by GAM and GLM from the analyses based on our expert opinion on the palm distributions to avoid introducing uncertainties into the ensemble framework. Hence, the ensemble predictions were only computed based on MAXENT and GBM.

For species with low sample sizes (30 > n ≥ 5; n = 26; Table A1), we used a bioclimatic envelope model called surface range envelope (SRE) as implemented in BIOMOD. For the SRE, two sets of models were tested: one using the complete predicted range per species (BIO00) and one confining the envelope between the 2.5 and 97.5 percentile range (BIO25). To avoid overfitting, we only used the first five principal components and the first five spatial filters for the SRE models. We decided to use species distributions predicted by BIO00 as opposed to BIO25, as BIO25 excluded well-documented occurrences and thereby underpredicted palm distributions.

A.4 Constructing continental species richness maps

In a final step, we constructed species richness maps by summing up the estimated distributions of the 62 palm species, i.e., the modeled distributions based on advanced SDM methods for species with ≥30 records (n = 25), the distributions derived from bioclimatic envelope models for species with low sample sizes between 5 and 30 records (n = 26), and the observed localities for the remaining species with <5 records (n = 11). To do this, the continuous occurrence probability from the GBM and MAXENT models were converted to binary predictions. We used the threshold that minimizes the difference between sensitivity and specificity which have been shown to perform amongst the best in comparative studies and to be superior to other threshold criteria when prevalence is low (Jiménez-Valverde and Lobo 2007) which is often the case for African palms at the continental scale (Table A1). We summed up the average species richness per cell predicted by the individual regression models (GBM and MAXENT) and that predicted by the SRE model, and finally added the per cell occurrences for species with <5 unique records at 100 × 100 km resolution. The reason for scaling up to 100 × 100 km for the richness analyses was due to the original resolution of the past climate data which is at 1° and 1.25° resolution, respectively (Table A4). Hence, finer scale analyses are not meaningful given the coarse resolution of the paleoclimate data.

Table A1. Overview of 62 species (95% of total species recognized for Africa) extracted from the African palm database giving information of unique occurrence records available for each species at 10 × 10 km resolution. The habitat affiliation and the estimated species' range size are given for each species. “Rainforest” includes species with strictly tropical rainforest affinity whereas “open-habitat” includes species that only occur in open habitats (covering savanna woodlands to dry deserts). Two species can occur in both habitats and are marked as “Mixed” and one species occurs in evergreen coastal forest and hence does not fall within either of the two categories.

Species

Unique occurrences

Habitat

Estimated range size (no. 10 ×10 km grids)

Borassus aethiopum

106

Open-habitat

47368

Borassus akeassii

13

Open-habitat

644

Calamus deerratus

140

Rainforest

36097

Chamaerops humilis

19

Open-habitat

1613

Dypsis pembana

2

Open-habitat

2

Elaeis guineensis

113

Mixed

34521

Eremospatha barendii

1

Rainforest

1

Eremospatha cabrae

33

Rainforest

18210

Eremospatha cuspidata

19

Rainforest

15293

Eremospatha dransfieldii

7

Rainforest

1261

Eremospatha haullevilleana

99

Rainforest

31093

Eremospatha hookeri

39

Rainforest

9281

Eremospatha laurentii

46

Rainforest

17706

Eremospatha macrocarpa

102

Rainforest

14068

Eremospatha quinquecostulata

7

Rainforest

174

Eremospatha tessmanniana

2

Rainforest

2

Eremospatha wendlandiana

45

Rainforest

6197

Hyphaene compressa

41

Open-habitat

8287

Hyphaene coriacea    

47

Open-habitat

6844

Hyphaene guineensis

18

Open-habitat

563

Hyphaene petersiana

66

Open-habitat

26199

Hyphaene reptans

3

Open-habitat

3

Hyphaene thebaica

48

Open-habitat

43130

Jubaeopsis caffra

2

Evergreen coastal forest

2

Laccosperma acutiflorum

25

Rainforest

5557

Laccosperma korupense

7

Rainforest

26

Laccosperma laeve    

39

Rainforest

6081

Laccosperma opacum

88

Rainforest

18423

Laccosperma robustum

53

Rainforest

15604

Laccosperma secundiflorum

99

Rainforest

22797

Livistona carinensis  

9

Open-habitat

19

Medemia argun          

14

Open-habitat

629

Oncocalamus macrospathus

14

Rainforest

4383

Oncocalamus mannii

30

Rainforest

7687

Oncocalamus tuleyi   

16

Rainforest

188

Oncocalamus wrightianus

5

Rainforest

28

Phoenix caespitosa    

7

Open-habitat

572

Phoenix reclinata      

294

Mixed

63118

Podococcus acaulis   

25

Rainforest

682

Podococcus barteri   

65

Rainforest

5509

Raphia africana         

2

Rainforest

2

Raphia australis         

5

Open-habitat

24

Raphia farinifera       

39

Open-habitat

23590

Raphia gentiliana      

9

Rainforest

9917

Raphia hookeri           

58

Rainforest

10736

Raphia laurentii         

8

Rainforest

638

Raphia longiflora      

1

Rainforest

1

Raphia mambillensis

13

Open-habitat

424

Raphia mannii            

1

Rainforest

1

Raphia matombe        

5

Rainforest

29

Raphia monbuttorum

7

Rainforest

1786

Raphia palma-pinus

32

Rainforest

8664

Raphia regalis            

23

Rainforest

2673

Raphia rostrata          

3

Rainforest

3

Raphia ruwenzorica  

3

Open-habitat

3

Raphia sese                 

10

Rainforest

2709

Raphia sudanica        

31

Open-habitat

12963

Raphia textilis            

2

Rainforest

2

Raphia vinifera          

13

Rainforest

597

Sclerosperma mannii

43

Rainforest

15524

Sclerosperma profizianum

11

Rainforest

5213

Sclerosperma walkeri

11

Rainforest

4224

 

Table A2. Wilcoxon signed-rank test for comparison of the area under the receiver operating curve (AUC) and the true skill statistic (TSS), both means of discriminatory ability among the species distribution models for 25 African palm species with ≥ 30 unique occurrence records at 10 × 10 km resolution.

Algorithms

AUC

TSS

Median (mean)

[min-max]

Median (mean)

[min-max]

Maxent

0.968 (0.964)a

0.858-0.997

0.896 (0.886)a

0.677-0.990

GLM

0.964 (0.947)b

0.781-0.996

0.891 (0.858)b

0.578-0.990

GAM

0.957 (0.944)b

0.774-0.990

0.874 (0.845)b

0.564-0.977

GBM

0.940 (0.939)b

0.857-0.996

0.867 (0.849)b

0.635-0.985

GLM, Generalized linear model; GAM, Generalized additive model (using 4 splines); GBM, Generalized boosting model.
Different superscript letters (a–b) indicate models which are significantly different at P < 0.001 except for MAXENT vs. GBM (TSS) and MAXENT vs. GLM (TSS) at P = 0.002 and P = 0.004, respectively.

 

Table A3. Goodness of fit-(G-)test for whether the observed frequency of the best performing species distribution model within BIOMOD (GBM, GLM, and GAM) based on area under the receiver operating curve (AUC) and the true skill statistics (TSS) deviates from random expectation (expected frequency = n × (1/total number of models), n = 50). Only one test is shown given that the observed frequencies are the same for AUC and TSS. The G-test was implemented using William's correction to get a better approximation to χ² (Gadj).

Models

Observed

Expected

Gadj

GBM

47

16.67

86.00***

GAM

3

16.67

 

GLM

0

16.67

 

GBM, Generalized Boosting Model; GAM, Generalized Additive Model; GLM, Generalized Linear Model. ***, p < 0.001.

 

Table A4. Predictor variables used to explain spatial variation in palm species richness across Africa. Original data were projected to the Lambert Azimuthal Equal Area projection and mean values were extracted for each grid cell (except for TOPO, SOIL, and VEG) at 100 × 100 km resolution.

Acronym

Predictor (units)

Computations and details

Original resolution and data source

Palaeoclimate

LGMPREC

LGM precipitation anomaly (mm/yr)

The ensemble mean of precipitation and temperature projected by two different Global Climate Models (Model for Interdisciplinary Research on Climate version 3.2 [MIROC 3.2] and the Community Climate System Model version 3 [CCSM3]) were used for the LGM. For all time periods, the anomalies were computed by: (1) resampling the past climate layers to the resolution of the contemporary climate data by bilinear interpolation techniques, and (2) using the following calculation: past climate minus present-day climate.

2.5' - The PMIP 2 project downloaded from Worldclim (Braconnot et al. 2007).

LGMTEMP

LGM temperature anomaly (°C)

PLIOPREC

Pliocene precipitation anomaly (mm/yr)

1° - (Haywood and Valdes 2004)

PLIOTEMP

Pliocene temperature anomaly (°C)

MIOPREC

Miocene precipitation anomaly (mm/yr)

1.25° - (Pound et al. 2011)

MIOTEMP

Miocene temperature anomaly (°C)

Current climate

PREC

Annual precipitation (in mm/yr)

 

5' - The Worldclim dataset (Hijmans et al. 2005)

MAT

Annual mean temperature (in °C)

 

AET

Actual evapotranspiration (in mm/yr)

 

10' - GNV183 data (Ahn and Tateishi 1994)

PET

Potential evapotranspiration (in mm/yr)

 

NDVI

Normalized difference vegetation index

The annual mean normalized difference vegetation index (NDVI) composed of monthly NOAA-AVHRR satellite images averaged over an 18 year period (1982-2000 excluding 1994) as a proxy for net primary productivity

10' - (EDIT Geoplatform 2010)

Habitat heterogeneity

TOPO

Topographic range (in meters)

Difference between maximum and minimum elevation extracted within each 100 × 100-km grid cell

1 km - U.S. geological Survey GTOPO30 digital elevation model (USGS 2008)

SOIL

Soil heterogeneity (number of soil types per grid)

The sum of soil types based on presences of each type extracted within each 100 × 100-km grid cell*

30'' - Harmonised World Soil Database ((FAO/IIASA/ISRIC/ISSCAS/JRC 2008)

VEG

Vegetation type heterogeneity (number of vegetation types per grid)

The sum of vegetation types  based on presences of each type extracted within each 100 × 100-km grid cell*

300 meter - GlobCover (ESA and
MEDIAS-France 2009)

Anthropogenic impact

HII

Human influence index (unit less)

HII consists of values ranging for Africa from low (0) to high (64) human influence, estimated per grid cell at 1 × 1 km based on several human impact proxy layers such as population density, land transformations, electric power infrastructure and accessibility from roads, rivers and coastlines

30'' - (Sanderson et al. 2002)

HPH

Mean historical human population density (in person km-2)

Human population density averaged over year 0AD, 500AD, 1000AD, and 1500AD

5' - HYDE 3.1 spatially explicit database (Klein Goldewijk et al. 2011)

*Using the Thematic Summary Raster in the ArcGIS extension Hawth Tool (version 3.26). LGM, Last glacial maximum (21,000 years ago).


Table A5. Pairwise Spearman's rank correlations (ρ) of present-day precipitation and temperature with late Miocene, Pliocene and Last Glacial Maximum (LGM) precipitation and temperature.

 

Present-day

 

Precipitation

Temperature

LGM

0.959

0.974

Pliocene

0.952

0.876

Miocene

0.710

0.656

 

Literature cited

Ahn, C.-H., and R. Tateishi. 1994. Development of global 30-minute grid potential evapotranspiration data set. Journal of the Japan Society of Photogrammetry and Remote Sensing 33:12–21.

Allouche, O., A. Tsoar, and R. Kadmon. 2006. Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). Journal of Applied Ecology 43:1223–1232.

Ali, M. 2004. Aquatic and shoreline vegetation of Lake Nubia, Sudan. Acta Botanica Croatica 63:101–111.

Alonso, L. E., F. Lauginie, and G. Rondeau, editors. 2005. A rapid biological assessment of two classified forests in south-western Côte d'Ivoire. RAP Bulletin of Biological Assessment No. 34. Conservation International, Washington, DC, USA.

Amwatta, C. J. M. 2004. Diversity of use of Doum Palm (Hyphaene compressa) leaves in Kenya. Palms 48:184–190.

Araújo, M. B., and M. New. 2007. Ensemble forecasting of species distributions. Trends in Ecology and Evolution 22:42–47.

Baranga, D. 2004. Forest fragmentation and primates' survival status in non-reserved forests of the 'Kampala area', Uganda. African Journal of Ecology 42:70–77.

Barbier, C. 1985. Further notes on Livistona carinensis in Somalia. Principes 29:151–155.

Barot, S., and J. Gignoux. 1999. Population structure and life cycle of Borassus aethiopum Mart.: evidence of early senescence in a palm tree. Biotropica 31:439–448.

Bayton, R. P., C. Obunyali, and R. Ranaivojaona. 2003. A Re-examination of Borassus in Madagascar. PALMS 47:206–219.

Bayton, R. P., A. Ouédraogo, and S. Guinko. 2006. The genus Borassus (Arecaceae) in West Africa, with a description of a new species from Burkina Faso. Botanical Journal of the Linnean Society 150:419–427.

Bayton, R. P., and A. Ouédraogo. 2009. Discovering Africa's newest palm. PALMS 53:37–45.

Beentje, H. J. 1990. Botanical assessment of Ngezi forest, Pemba. Prepared for the Zanzibar Forestry Development Project of FINNIDA and the Finnish National Board of Forestry. 

Blach-Overgaard A., J.-C. Svenning, and H. Balslev. 2009. Climate change sensitivity of the African ivory nut palm, Hyphaene petersiana Klotzsch ex Mart. (Arecaceae) - a keystone species in SE Africa. IOP Conference Series: Earth and Environmental Science 8:012014.

Blach-Overgaard, A., J.-C. Svenning, J. Dransfield, M. Greve, and H. Balslev. 2010. Determinants of palm species distribution across Africa: the relative roles of climate, non-climatic environmental factors, and spatial constraints. Ecography 33:380–391.

Borcard, D., and P. Legendre. 2002. All-scale spatial analysis of ecological data by means of principal coordinates of neighbour matrices. Ecological Modelling 153:51–68.

Bornkamm, R., I. Springuel, F. Darius, M. G. Sheded, and M. Radi. 2000. Some observations on the plant communities of Dungul Oasis (Western Desert, Egypt). Acta Botanica Croatica 59:101–109.

Braconnot, P., B. Otto-Bliesner, S. Harrison, S. Joussaume, J.-Y. Peterchmitt, A. Abe-Ouchi, M. Crucifix, E. Driesschaert, T. Fichefet, C. D. Hewitt, M. Kageyama, A. Kitoh, A. Laîné, M.-F. Loutre, O. Marti, U. Merkel, G. Ramstein, P. Valdes, S. L. Weber, Y. Yu, and Y. Zhao. 2007. Results of PMIP2 coupled simulations of the Mid-Holocene and Last Glacial Maximum - Part 1: experiments and large-scale features. Climate of the Past 3:261–277.

Brown, E. 1991. The Quest for Jubaeopsis caffra, the Pondoland Palm. Principes 35:99–101.

Bullock, S. H. 1980. Demography of an undergrowth palm in littoral Cameroon. Biotropica 12:247–255.

CIESIN (Center for International Earth Science Information Network) and CIAT (Columbia University; and Centro Internacional de Agricultura Tropical). 2005. Gridded population of the World version 3 (GPWv3). Palisades, NY: Socioeconomic Data and Applications Center (SEDAC), Columbia University. Available at http://sedac.ciesin.columbia.edu/gpw. Last assessed 24 April 2008.

Corkill, N. L. 1949. Dietary change in a Sudan village following locust visitation. Africa 19:1–12.

Cunningham, A. B., and S. J. Milton. 1987. Effects of basket-weaving industry on Mokola palm and dye plants in northwestern Botswana. Economic Botany 41:386–402.

De Marco P., J. A. F. Diniz-Filho, and L. M. Bini. 2008. Spatial analysis improves species distribution modelling during range expansion. Biology Letters 4:577–580.

Dransfield, J. 1986. Palmae. Pages 1–55 in R. M. Polhill, editor. Flora of tropical East Africa. A.A. Balkema, Rotterdam, The Netherlands.

Dransfield, J. 1988. The palms of Africa and their relationships. Pages 95–103 in P. Goldblatt and P.P. Lowry, editors. Modern Systematic Studies in African Botany. Missouri Botanical Garden Press, St. Louise, USA.

Dransfield, J., N. W. Uhl, C. B. Asmussen et al. 2008a. Genera Palmarum. The evolution and classification of palms. Royal Botanic Gardens, Kew, Richmond, UK.

Dransfield, J., M. Rakotoarinivo, W. J. Baker, R. P. Bayton, J. B. Fisher, J. W. Horn, B. Leroy, and X. Metz. 2008b. A new Coryphoid palm genus from Madagascar. Botanical Journal of the Linnean Society 156:79–91.

EDIT Geoplatform. 2010. Normalized difference vegetation index (NDVI). Available at: http://edit.csic.es/Soil-Vegetation-LandCover.html. Last accessed 10 March 2010.

Eilu, G., D. L. N. Hafashimana, and J. M. Kasenene. 2004. Tree species distribution in forests of the Albertine Rift, western Uganda. African Journal of Ecology 42:100–110.

Elith, J. 2000. Quantitative methods for modeling species habitat: comparative performance and an application to Australian plants. Pages 39–58 in Ferson S. and Burgman M.A., editors. Quantitative methods in conservation biology. Springer-Verlag, New York, New York, USA.

Elith, J., and C. H. Graham. 2009. Do they? How do they? WHY do they differ? On finding reasons for differing performances of species distribution models. Ecography 32:66–77.

Elith, J., H. Graham, P. Anderson, M. Dudik, S. Ferrier, A. Guisan, J. Hijmans, F. Huettmann, R. Leathwick, A. Lehmann, J. Li, G. Lohmann, A. Loiselle, G. Manion, C. Moritz, M. Nakamura, Y. Nakazawa, C. M. Overton, A. Townsend Peterson, J. Phillips, K. Richardson, R. Scachetti-Pereira, E. Schapire, J. Soberon, S. Williams, S. Wisz, and E. Zimmermann. 2006. Novel methods improve prediction of species' distributions from occurrence data. Ecography 29:129–151.

Elith, J., M. Kearney, and S. Phillips. 2010. The art of modelling range-shifting species. Methods in Ecology and Evolution 1:330–342.

ESA and MEDIAS-France. 2009. GlobCover Land Cover version 2.2. Available at: http://www.esa.int/due/ionia/globcover. Last accessed 10 March 2010.

FAO/IIASA/ISRIC/ISSCAS/JRC. 2008. Harmonized World Soil Database (version 1.0). FAO, Rome, Italy and IIASA, Laxenburg, Austria. Available at: http://www.iiasa.ac.at/Research/LUC/External-World-soil-database/html. Last accessed 16 October 2008.

Ford, H., and C. Bealy. 2004. Status of the Bankoualé Palm, Livistona carinensis, in Djibouti. PALMS 48:94–101.

Fujioka, Y. 2005. Vegetation changes and use of palms as a building material by Ovambo agro-pastoralists in north-central Namibia. African Study Monographs, Suppl. 30:89–105.

Gibbons, M., and T. W. Spanner. 1996. Medemia argun lives! Principes 40:65–74.

Govaerts, R., J. Dransfield , S. F. Zona, D. R. Hodel and A. Henderson. 2013. World Checklist of Arecaceae. Facilitated by the Royal Botanic Gardens Kew, UK: Available at: http://apps.kew.org/wcsp/. (Last accessed 8 March 2013).

Griffith, D. A. 2003. Spatial autocorrelation and spatial filtering: gaining understanding through theory and scientific visualization. Springer-Verlag, Berlin, Germany.

Guisan, A., and W. Thuiller. 2005. Predicting species distribution: offering more than simple habitat models. Ecology Letters 8:993–1009.

Guisan, A., and N. E. Zimmermann. 2000. Predictive habitat distribution models in ecology. Ecological Modelling 135:147–186.

Hansen, M. C., R. S. DeFries, J. R. G. Townshend, R. Sohlberg, C. Dimiceli, and M. Carroll. 2002. Towards an operational MODIS continuous field of percent tree cover algorithm: examples using AVHRR and MODIS data. Remote Sensing of Environment 83:303–319.

Haywood, A. M., and P. J. Valdes. 2004. Modelling Pliocene warmth: contribution of atmosphere, oceans and cryosphere. Earth and Planetary Science Letters 218:363–377.

Hijmans, R. J., S. E. Cameron, J. L. Parra, P. G. Jones, and A. Jarvis. 2005. Very high resolution interpolated climate surfaces for global land areas. International Journal of Climatology 25:1965–1978.

Hoke, P., R. Demey and A. Peal, editors. 2007. A rapid biological assessment of North Lorma, Gola and Grebo National Forests, Liberia. RAP Bulletin of Biological Assessment 44. Conservation International, Arlington, VA, USA.

Hovestadt, T., P. Yao, and K. E. Linsenmair. 1999. Seed dispersal mechanisms and the vegetation of forest islands in a West African forest-savanna mosaic (Comoé National Park, Ivory Coast). Plant Ecology 144:1–25.

Ibrahim, H., and W. J. Baker. 2009. Medemia argun - past, present and future. PALMS 53:9–21.

Jiménez-Valverde, A., and J. M. Lobo. 2007. Threshold criteria for conversion of probability of species presence to either-or presence-absence. Acta Oecologica 31:361–369.

Johnson, D. V. 1984. Notes on the Palms of Guinea-Bissau. Principes 28:155–162.

Johnson, D. V. 1991. The Mpapindi Palm (Chrysalidocarpus pembanus) of Pemba Island, Tanzania. Principes 35:83–85.

Kakudidi, E. K. 2004. Cultural and social uses of plants from and around Kibale National Park, Western Uganda. African Journal of Ecology 42:114–118.

Kgathi, D. L., G. Mmopelwa, and K. Mosepele. 2005. Natural resources assessment in the Okavango Delta, Botswana: Case studies of some key resources. Natural Resources Forum 29:70–81.

Kinnaird, M. F. 1992. Competition for a forest palm - use of Phoenix reclinata by human and nonhuman primates. Conservation Biology 6:101–107.

Klein Goldewijk, K., A. Beusen, G. van Drecht, and M. de Vos. 2011. The HYDE 3.1 spatially explicit database of human-induced global land-use change over the past 12,000 years. Global Ecology and Biogeography 20:73–86.

Kolongo, T. S. D., G. Decocq, C. Yao, E. Blom, and R. Van Rompaey. 2006. Plant species diversity in the southern part of the Taï National Park (Côte d'Ivoire). Biodiversity and Conservation 15:2123–2142.

Konstant, T., S. Sullivan, and A. Cunningham. 1995. The effects of utilization by people and livestock on Hyphaene petersiana (Arecaceae) basketry resources in the palm savanna of North-Central Namibia. Economic Botany 49:345–356.

Kouassi, K. I., S. Barot, and I. A. Z. Bi. 2009. Population structure and reproductive strategy of two multiple-stemmed rattans of Cote d'Ivoire. PALMS 53:20–30.

Kouassi, K. I., S. Barot, J. Gignoux, and I. A. Z. Bi. 2008. Demography and life history of two rattan species, Eremospatha macrocarpa and Laccosperma secundiflorum, in Cote d'Ivoire. Journal of Tropical Ecology 24:493–503.

Lambert, J. 1998. Primate frugivory in Kibale National Park, Uganda, and its implications for human use of forest resources. African Journal of Ecology 36:234–240.

Lobo, J. M., A. Jimenez-Valverde, and R. Real. 2008. AUC: a misleading measure of the performance of predictive distribution models. Global Ecology and Biogeography 17:145–151.

Long, D. G., M. R. Drinkwater, B. Holt, S. Saatchi, and C. Bertoia. 2001. Global ice and land climate studies using scatterometer image data. EOS Transactions AGU 82:503.

Markham, R. H., and A. J. Babbedge. 1979. Soil and vegetation catenas on the forest-savanna boundary in Ghana. Biotropica 11:224–234.

McCullough, J., editor. 2004. A rapid biological assessment of the Forêt Classée du Pic de Fon, Simandou Range, Southeastern Republic of Guinea. RAP Bulletin of Biological Assessment 35. Conservation International, Washington, DC, USA.

McCullough, J., Decher, J. Kpelle D. Guba, editors. 2005. A biological assessment of the terrestrial ecosystems of the Draw River, Boi-Tano, Tano Nimiri and Krokosua Hills forest reserves, southwestern Ghana. RAP Bulletin of Biological Assessment 36.  Conservation International, Washington, DC, USA.

McCullough, J., L. E. Alonso, P. Naskrecki, H. E. Wright, and Y. Osei-Owusu, editors. 2007. A rapid biological assessment of the Atewa Range Forest Reserve, Eastern Ghana. RAP Bulletin of Biological Assessment 47.  Conservation International, Washington, DC, USA.

Medley, K. 1993. Extractive forest resources of the Tana River national primate reserve, Kenya. Economic Botany 47:171–183.

Moll, E. J. 1972. The distribution, abundance and utilization of the Lala Palm, Hyphaene natalensis, in Tongaland, Natal. Bothalia 10:627–636.

Mollet, M., F. Herzog, Y. E. N. Behi, and Z. Farah. 2000. Sustainable exploitation of Borassus aethiopum, Elaeis guineensis and Raphia hookeri for the extraction of palm wine in Côte d'Ivoire. Environment, Development and Sustainability 2:45–59.

Mwaura, F., and H. M. Kaburu. 2009. Spatial variability in woody species richness along altitudinal gradient in a lowland-dryland site, Lokapel Turkana, kenya. Biodiversity and Conservation 18:19–32.

Nangendo, G., H. Steege, and F. Bongers. 2006. Composition of woody species in a dynamic forest-woodland-savannah mosaic in Uganda: implications for conservation and management. Biodiversity and Conservation 15:1467–1495.

New, M., D. Lister, M. Hulme, and I. Makin. 2002. A high-resolution data set of surface climate over global land areas. Climate Research 21:1–25.

Newton-Fisher, N. E. 1999. The diet of chimpanzees in the Budongo Forest Reserve, Uganda. African Journal of Ecology 37:344–354.

Obermeyer, A. A., and R. G. Strey. 1969. A new species of Raphia from northern Zululand and southern Mosambique. Bothalia 10:29–37.

Ogunjemite, B. G., T. A. Afolayan, and E. A. Agbelusi. 2005. Habitat structure of chimpanzee community in Ise-Forest Reserve, Ekiti State, South-western Nigeria. African Journal of Ecology 43:396–399.

Okolo, E. C. 1988. Chromosome counts on Nigerian species of the genus Raphia. Principes 32:156–159.

Otedoh, M. O. 1975. The distribution of Raphia regalis Becc. (Palmae) in Africa. The Nigerian Field 40:172–178.

Otedoh, M. O. 1982. A revision of the genus Raphia Beauv. (Palmae). Journal of the Nigerian Institue for Oil Palm Research 6:145–189.

Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190:231–259.

Phillips, S. J. and M. Dudik. 2008. Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31:161–175.

Pound, M. J., A. M. Haywood, U. Salzmann, J. B. Riding, D. J. Lunt, and S. J. Hunter. 2011. A Tortonian (Late Miocene, 11.61-7.25 Ma) global vegetation reconstruction. Palaeogeography, Palaeoclimatology, Palaeoecology 300:29–45.

Profizi, J.-P. 1985. Raphia hookeri: a survey of some aspects of the growth of a useful swamp Lepidocaryoid palm in Benin (West Africa). Principes 29:108–114.

R Development Core Team. R: a language and environment for statistical computing. [2.11.1]. 2010. Vienna, R foundation for statistical computing.

Rangel, T. F. L. V. B., J. A. F. Diniz-Filho, and L. M. Bini. 2006. Towards an integrated computational tool for spatial analysis in macroecology and biogeography. Global Ecology and Biogeography 15:321–327.

Russell, T. A. 1968. Palmae. Pages 159–169 in F. N. Hepper, editor. Flora of west tropical Africa. The Whitefriars Press Ltd., London, UK.

Saatchi, S., W. Buermann, H. Ter Steege, S. Mori, and T. B. Smith. 2008. Modeling distribution of Amazonian tree species and diversity using remote sensing measurements. Remote Sensing of Environment 112:2000–2017.

Sambou, B., A. Goudiaby, F. Ervik, D. Diallo, and M. C. Camara. 2002. Palm wine harvesting by the Bassari threatens Borassus aethiopum populations in north-western Guinea. Biodiversity and Conservation 11:1149–1161.

Sanderson, E. W., M. Jaiteh, M. A. Levy, K. H. Redford, A. V. Wannebo, and G. Woolmer. 2002. The human footprint and the last of the wild. Bioscience 52:891–904.

Schulenberg, T. S, C. A. Short, and P. J. Stephenson, editors. 1999. A biological assessment of Parc National de la Marahoué, Côte d'Ivoire. RAP Working Papers 13. Conservation International, Washington, DC, USA.

Shapcott, A., J. Dowe, and H. Ford. 2009. Low genetic diversity and recovery implications of the vulnerable Bankoualé palm Livistona carinensis (Arecaceae), from North-eastern Africa and the Southern Arabian Peninsula. Conservation Genetics 10:317–327.

Sneed, M. W. 1983. Hyphaene petersiana amongst animals in the Heartland of Africa. Principes 27:143–151.

Sokal, R. R., and F. J. Rohlf. 1995. Biometry. W. H. Freeman and Company, New York, New York, USA.

Stave, J., G. Oba, C. S. Bjora, Z. Mengistu, I. Nordal, and N. C. Stenseth. 2003. Spatial and temporal woodland patterns along the lower Turkwel River, Kenya. African Journal of Ecology 41:224–236.

Stave, J., G. Oba, I. Nordal, and N. C. Stenseth. 2006. Seedling establishment of Acacia tortilis and Hyphaene compressa in the Turkwel riverine forest, Kenya. African Journal of Ecology 44:178–185.

Sugiyama, Y. 1985. The brush-stick of chimpanzees found in south-west Cameroon and their cultural characteristics. PRIMATES 26:361–374.

Sunderland, T. C. H. 2003. Two new species of Rattan (Palmae: Calamoideae) from the forests of West and Central Africa. Kew Bulletin 58:987–990.

Sunderland, T. 2007. Field guide to the Rattans of Africa, Royal Botanic Gardens, Kew, UK.

Thuiller, W., B. Lafourcade, R. Engler, and M. B. Araujo. 2009. BIOMOD - a platform for ensemble forecasting of species distributions. Ecography 32:369–373.

Tuley, P. 1995. The palms of Africa, The Trendrine Press. Zennor, St. Ives, Cornwall, UK.

USGS. 2008. Hydro1K Africa. Available at: http://eros.usgs.gov/#/Find_Data/Products_and_Data_Available/gtopo30/hydro/africa. Last accessed 24 April 2008.

van Gemerden, B. S., H. Olff, M. P. E. Parren, and F. Bongers. 2003. The pristine rain forest? Remnants of historical human impacts on current tree species composition and diversity. Journal of Biogeography 30:1381–1390.

van Valkenburg, J., and J. Dransfield. 2004. Hyphaene guineensis. PALMS 48:10–16.

van Valkenburg, J., and T. Sunderland. 2008. A revision of the genus Podococcus ( Arecaceae ). Kew Bulletin 63:251–260.

van Valkenburg, J., T. Sunderland, and T. Couvreur. 2008. A revision of the genus Sclerosperma (Arecaceae ). Kew Bulletin 63:75–86.

Wahungu, G. E. O. F. 1998. Diet and habitat overlap in two sympatric primate species, the Tana crested mangabey Cercocebus galeritus and yellow baboon Papio cynocephalus. African Journal of Ecology 36:159–173.

Wicht, H. 1969. The indigenous palms of southern Africa, Citadel Press, Lansdowne, South Africa.

Wisz, M., and A. Guisan. 2009. Do pseudo-absence selection strategies influence species distribution models and their predictions? An information-theoretic approach based on simulated data. BMC Ecology 9:8.

Wong, S., T. Saj, and P. Sicotte. 2006. Comparison of habitat quality and diet of Colobus vellerosus in forest fragments in Ghana. Primates 47:365–373.

Wright, H. E., J. McCullough, L. E. Alonso, and M. S. Diallo, editors. 2006a. A rapid biological assessment of three classified forests in southeastern Guinea. RAP Bulletin of Biological Assessment 40. Conservation International, Washington, DC, USA.

Wright, H. E., J. McCullough, and M. S. Diallo,editors. 2006b. A rapid biological assessment of the Boké Préfecture, Northwestern Guinea. RAP Bulletin of Biological Assessment 41. Conservation International, Washington, DC, USA.

Yamakoshi, G. 1998. Dietary responses to fruit scarcity of wild chimpanzees at Bossou, Guinea: Possible implications for ecological importance of tool use. American Journal of Physical Anthropology 106:283–295.


[Back to E094-222]