Topographic Controls on the Distribution of Tree Islands in the High Andes of South-western Ecuador

8/8/2019 Topographic Controls on the Distribution of Tree Islands in the High Andes of South-western Ecuador

1/13

O R I G I N A LA R T I C L E

Topographic controls on the distribution

of tree islands in the high Andes of

south-western Ecuador

D. Coblentz1* and P. L. Keating2

INT RO DUCT IO N

Explaining the spatial distribution of plant communities has

long been a goal of ecologists (e.g. Forman & Godron, 1986;

Turner et al., 2001). A long history of investigations has

established that the nature of the topographic fabric, typically

evaluated through geomorphometry, or the science concerned

with the geometry of the landscape (Chorley et al., 1957),

1Los Alamos National Laboratory, Los Alamos,

NM, USA and2Department of Geography,

Indiana University, Bloomington, IN, USA

*Correspondence: David Coblentz, MS F665,

Los Alamos National Laboratory, Los Alamos,

NM 87545, USA.

E-mail: [email protected]

ABS T RACT

Aim To evaluate the hypothesis that geomorphometric parameters of upper

montane Andean environments have an important influence on the regional fire

ecology and consequently play a role in the spatial distribution of remnant tree

islands dominated by Polylepis.

Location A glacial landscape located between 3600 and 4400 m elevation in

Cajas National Park, south-western Ecuador.

Methods The eigenvalue ratio method was used to evaluate the regional

geomorphometric parameters of a 30-m digital elevation model for CajasNational Park. The landscape character was evaluated by quantifying the

topographic roughness, organization, and gradient. This information was used to

determine the spatial correlations between terrain characteristics and the

distribution of tree islands in the region.

Results We demonstrate a strong spatial correlation between areas of high

topographic roughness and gradient, and the locations of the major tree islands.

We find that there is a distinctive relationship between the topographic roughness

and organization in the vicinity of the tree islands (e.g. increased upslope

roughness and decreased topographic grain strength) that substantiates the

notion that the tree islands are located in relatively inaccessible topography.

Main conclusions In the northern and central Andes, the location ofPolylepis-dominated forest islands has been shown to be a function of climate, terrain

characteristics, and anthropogenic disturbances. Although the relative importance

of various ecological factors has been debated, it remains clear that fires have

exerted a strong influence on these ecosystems. Other authors have noted that

tree islands are more likely to occur at the base of cliffs, above moist areas, and in

other areas where fires do not burn frequently. Our results corroborate these

observations, and demonstrate that the occurrence ofPolylepis patches is strongly

correlated with specific combinations of terrain features. Although we do not

discount the importance of other factors in determining the spatial position and

areal extent of these forests, we demonstrate strong support for fire-related

hypotheses.

Keywords

Andes, Cajas National Park, fire ecology, geoecology, geomorphometry,

landscape ecology, Polylepis, remote sensing, topographic analysis, tropical

montane forest.

Journal of Biogeography(J. Biogeogr.) (2008) 35, 20262038

2026 www.blackwellpublishing.com/jbi

2008 The Authorsdoi:10.1111/j.1365-2699.2008.01956.x Journal compilation 2008 Blackwell Publishing Ltd


2/13

plays a fundamental role in the distribution of vegetation and

biodiversity (Coblentz & Riitters, 2004). Despite the important

link between topography and geoecology in montane regions,

most studies of this relationship have been limited to a

primarilyqualitative framework (e.g. Brown & Gibson, 1983).

Digital elevations models (DEMs), rectangular grids of eleva-

tion values, provide an unparalleled tool for the evaluation of

geomorphometric parameters. The current availability of high-

speed computing platforms, high-resolution DEMs now glob-

ally available at a 3-arc-second (about 90 m) resolution, and

high-resolution land cover data sets (e.g. Riitters et al., 2000)make it possible to undertake a rigorous quantitative analysis

that can be used to test the relationships discussed in previous

qualitative investigations.

In the Andean highlands, the spatial distribution of tropical

forests has been shown to be a function of both topographic

features and anthropogenic disturbances (e.g. Troll, 1968,

1973; Young & Keating, 2001). Elevation establishes a complex

set of environmental gradients, including temperature regimes

and precipitation (Troll, 1968; Lauer, 1981). In addition,

upper montane ecosystems are strongly influenced by

topographic features such as slope, aspect, and shape (Troll,

1973; Young, 1998a; Keating, 1999). Given that humans have

been present in the Andes for at least 7000 years (Bruhns,

1994), disturbances have also influenced the composition,

structure and spatial position of forest communities. Agricul-

tural practices, burning, grazing, road construction, and

logging have continually modified many areas of the Andes

(Ellenberg, 1979; Lgaard, 1992; Young, 1994, 1998b; Keating,

1998), leading to a land cover that is characterized by a

complex mosaic of vegetation types (Gade, 1999). Under-

standing the timing and exact nature of landscape transfor-mations is difficult, as each region has a different landscape

history (Young & Zimmerer, 1998). Burning, however, con-

tinues to occur in conjunction with most land-use activities,

especially grazing, and is probably the most significant

disturbance in the Ecuadorian Andes above 3500 m elevation

(Keating, 2007).

At elevations above 3000 m, the Ecuadorian Andes are

covered with a spatially heterogeneous assemblage of agricul-

tural lands, paramo (tropical alpine) and forest communities.

Whereas continuous forest covers the outer flanks of both

Figure 1 Topographic map of south-wes-

tern Ecuador with the study area indicated by

the solid box. In the lower figure, the tree

islands of Cajas National Park are designated

by polygons with colours representing four

area designations: red corresponds to tree

islands covering an area < 1 ha; green to

15 ha; blue to 510 ha; and black to

> 10 ha.

Topographic controls on tree islands

Journal of Biogeography35, 20262038 2027 2008 The Authors. Journal compilation 2008 Blackwell Publishing Ltd


3/13

cordilleras up to 4300 m elevation, inter-Andean forest (sensu

Sarmiento, 1995) tends to occur as remnant forest patches in

remote locations (Valencia & Jrgensen, 1992; Stern, 1995;

Young & Keating, 2001). Some continuous forest occurs in the

southern tip of the country (Madsen & llgaard, 1994), but

intensive, long-term land use has reduced most Andean forests

to a series of strips and patches found in gullies, ravines,

volcanic calderas, and remote high-elevation sites. Fragments

of fire-sensitive vegetation are characteristic of many fire-

prone landscapes (Bowman, 2000), and topographic protec-

tion of remnant vegetation communities is well documented in

several tropical ecosystems (e.g. Fensham, 1995).

In many sections of the Andes, including south-western

Ecuador, high-elevation patches ofPolylepis-dominated forests

occur above the timberline separating continuous, closed

forest from paramo (Fig. 1). Polylepis is a taxonomically

complex genus comprising c. 20 species whose centre of

diversity occurs between central Peru and southern Bolivia

(Kessler, 1995; Navarro et al., 2005). Often found at up to

5000 m elevation in Bolivia, these trees occur at the highest

elevation of any angiosperm tree taxa (Navarro et al., 2005).Seven species of Polylepis are found in Ecuador, where they

occur between 2700 and 4300 m a.s.l. (Romoleroux, 1994).

Whereas Polylepis incana may occur as monospecific stands

(Fehse et al., 2002), most Polylepis forest islands in Ecuador

contain other tree taxa, including Columella, Escallonia and

Gynoxys (Fjeldsa, 1992; Romoleroux, 1994).

Several thousand years ago, Polylepis forests probably

covered a far more extensive area than they do at present.

Palaeohistorical evidence suggests that the geographical distri-

butions of these forests fluctuated widely, largely in response to

climatic changes (Simpson, 1979). As a result of the dramatic

landscape transformations described above, however, there has

been a substantial reduction in the area covered. Fires, grazing

and wood-cutting appear to be the principal factors explaining

this change, and these activities continue to restrict the spatial

extent of upper montane forests (e.g. Fjeldsa, 1992; Hensen,

2002; Enrico et al., 2004; Renison et al., 2006). Given the

frequency of these disturbances, these forests now occur

primarily in ravines, on steep rocky terrain, near streams, and

in other protected sites (Lgaard, 1992; Kessler, 1995). Recent

studies indicate that only the upper limits of Polylepis patches

are determined by harsh climatic conditions, which impede the

growth and regeneration of trees (Lgaard, 1992; Cierjacks

et al., 2007); under these conditions, certain sites afford the

seedlings protection from the wind and cold (e.g. Simpson,1986; Ibisch, 1993).

In this article we focus primarily on the notion that fires play

a key role in maintaining the current spatial distribution of

forest patches. Our study is motivated by the hypothesis that

the tree islands in Cajas National Park are located in areas that

experience lower fire frequencies owing to local topographic

variations such as steep terrain. Other issues, such as climatic

factors and disturbance history, are beyond the scope of this

study. To test our hypothesis, we apply the eigenvalue ratio

method (Woodcock, 1977) to quantify the character of the

landscape and to evaluate the terrain where Polylepis patches

occur in a section of south-western Ecuador, and discuss

several salient trends in relation to the presence of fire. As

discussed below, this technique extracts several types of

information about the topographic fabric of the region, and

provides a comprehensive topographic analysis.

M AT E RIALS AND M E T HO DS

Study area

El Cajas National Park covers c. 29,000 hectares of upper

montane ecosystems on the western cordillera of the Ecuado-

rian Andes (Ministerio de Ambiente, 2007). Located c. 35 km

west of Cuenca, the middle of the park is found at 2 50 S,

7915 W (Fig. 1). Elevations within the park range from

c. 3180 to 4500 m. With the area having been glaciated during

the Pleistocene (Hastenrath, 1981; Clapperton, 1993), the

physiography includes more than 230 glacial lakes, intercon-

nected ridges and peaks, and numerous broad hanging valleys.

Glaciers were absent during the Holocene, and little tephra orpyroclastic materials from volcanoes have fallen on this region

(Rodbell et al., 2002). Soils in El Cajas did not develop from

underlying rocks, but instead are composed of organic and

aeolian materials (Harden, 2007), and are often referred to as

black Andean soils, histosols, or hydric andisols.

The dominant land cover is paramo, a form of tropical

alpine vegetation. Much of the valley bottom is characterized

by bogs or moist grasslands, whereas the slopes are better

drained and covered with tussock grasses. Woody species

occur infrequently in the grasslands, and shrubs are typically

found in isolated patches on the valley floors. Plant commu-

nity patterns are relatively uniform in the tussock-grass

paramo but vary markedly along the valley bottoms (Keating

et al., 2002). The species richness of this ecosystem is probably

as high as in any paramo in the country, and its floristic

composition differs significantly from that of paramos located

to the north of this region (Sklenar & Jrgensen, 1999; Sklenar

& Ramsay, 2001).

Three varieties of forest are found in the park. Continuous,

relatively undisturbed forest occurs in the lower sections of the

park; timberlines usually occur between 3330 and 3365 m a.s.l.

Second, a diverse mixture of secondary forest and tall shrubs

occurs near roads and along some valley bottoms. Finally,

forest patches, which often occur at the base of cliffs or steep

slopes (Fig. 2), are typically found above 3600 m a.s.l. Theseforest islands are often dominated by 510 m tall Polylepis

trees, although the interior portions may contain additional

tree genera, including Clethra, Gynoxys, and Myrsine. Generally

speaking, the relative importance of Polylepis increases above

3800 m (Hansen et al., 2003).

Climatic data have not been collected extensively for this

section of El Cajas, but it has been suggested that the paramos

in this region receive at least 2500 mm of precipitation

annually (Morris, 1985). Most of this falls between January and

June, and a relatively dry season occurs during the rest of the

D. Coblentz and P. L. Keating

2028 Journal of Biogeography 35, 20262038 2008 The Authors. Journal compilation 2008 Blackwell Publishing Ltd


4/13

year (Blandn, 1989). Maximum daily temperatures above

3500 m are usually between 3 and 7C when clouds are present

(P.L. Keating, personal observation); minimum temperatures

are usually between 0 and 4C.

The long-term environmental history of these highlands has

been investigated only recently, and much of the disturbance

history is a matter of speculation. After the late glacial period,

which ended 11,000 yr bp, Polylepis increased in importance at

higher elevations. Charcoal evidence suggests that fires also

became more frequent at that time; however, it appears that

this frequency decreased somewhat around 4000 years ago

(Hansen et al., 2003). The role played by humans in modifying

the landscape, however, is not clear; nor is there any

documented spatial correlation between forest communities

and the occurrence of fires.

Although the Ecuadorian highlands have been inhabited by

various indigenous groups during the past several thousand

years, past land-use practices and associated disturbance

regimes in the Cuenca region have received little attention.

Several anthropologists have examined pre-Conquest settle-

ment patterns and material culture (e.g. Collier & Murra, 1943;

Bruhns et al., 1990), and other authors have described more

recent settlement patterns in the region (e.g. Borrero, 1989).

To our knowledge, however, no one has provided detailed

descriptions of human impacts within El Cajas for any periodbefore the 20th century. It is worth noting, however, that early

Spanish explorers described these mountains as lacking forests

in the 1580s (Jimenez de Espada, 1965[1582]). We were unable

to find photographs or other records that document forest

cover change at any time during the past four centuries.

Between 1979 and 1996, this park was referred to as Las

Cajas Recreational Area (Fundacion Natura, 1992). Because

the designation of recreational areas in Ecuador does not

entail strict conservation protection, both burning and grazing

by livestock have precluded the regeneration of woody species

throughout the region. During the dry season, several fires

could occur simultaneously during any given day (Keating

et al., 2002). Many of the burns cover more than several

hundred hectares, so that the paramo represents various states

of recovery after these disturbances. Clearly, forest could not

regenerate under such conditions, and shrub paramo (sensu

Cuatrecasas, 1968), which often serves as a transition com-

munity between forests and grass paramo, rarely occurs. Given

that this region has been under the influence of human activity

for at least 3500 years (White & Maldonado, 1991), it is a

reasonable assumption to treat the forest patches within El

Cajas as relatively stable units. As we illustrate below, the

position and areal extent of these forest islands have probably

not changed during the past several decades. Only in rare cases

did forest margins show any signs of recent burning.

Analysis of vegetative cover

The vegetative cover in the study area was quantified through

an analysis of a 15 15 km subset of a Landsat Thematic

Mapper (TM) image (30-m cell size) acquired on 26 March1987. Although more recent imaging would be desirable,

Landsat, SPOT and Aster images collected subsequent to this

date have included substantially more cloud coverage and thus

are of limited utility for mapping land cover. Atmospheric

correction was performed with a technique described in Colby

& Keating (1998). The subscene was rectified using 23 ground

control points, and a root mean squared error (RMSE) of

0.230 was obtained. A nearest neighbour resampling algorithm

was used, and the rectified image was subset to the coordinates

given below. The normalized difference vegetation index

(NDVI) was calculated to distinguish the forested areas from

other land-cover types. Calculated as near infrared ) red/near

infrared + red, or TM bands 4 ) 3/4 + 3, this index provides

an indication of leafy green biomass (Sader et al., 1989; Mausel

et al., 1993). Because the forested areas contained significantly

more leafy biomass than the paramo vegetation, a density slice

technique (Jensen, 2005) was utilized to map the forest islands

and secondary forest. After the range of NDVI values was

determined for each cover type, a threshold was established to

separate trees from other land-cover types; aerial photographs

acquired on 9 October 1976 were used to assist in this

classification process.

To determine the characteristics and diversity of vegetation

that occurs within Cajas National Park, extensive field

reconnaissance was performed during June 1990 and October1991. Initially, a 9 12 km2 area was selected that included

representative physiographic features, numerous forest

patches, and several hundred hectares of secondary

forest located near roads and in valleys; continuous, primary

forest was not included in the study area. UTM coordinates

(Zone 17 S) for the upper left corner of the study site are

695000 E, 9696000 N, and those for the lower right corner are

707000 E, 9687000 N. Further ground verification was per-

formed during June and July 1995, and aerial photographs

acquired on 12 December 1995 were also used to check the

Figure 2 A typical tree island located in the south-west portion

of the study area at c. 3900 m in Cajas National Park. Dominated

by Polylepis, the forest patch covers c. 10 ha and is located in a

boulder-strewn area at the base of a cliff. Note that the tussock-

grass paramo surrounding this site has been burned within the

past year.




5/13

accuracy of the mapping process. Comparison of photographs

from both dates revealed that little, if any, change in forest

cover has occurred during the past 20 years. Approximately

15% of the scene was covered with clouds or cloud shadows,

and these areas were not included in the analysis.

Evaluation of geomorphometric parameters

Geomorphometry, the study of landscape geometry, provides a

way to quantify the relationships between topographic vari-

ables and vegetative land cover. Within biogeography, there

has been a growing recognition that topographic parameters

exert a strong influence on patterns of vegetative cover (see

discussion in Coblentz & Riitters, 2004), and are valuable in

wildlife habitat modelling (Sappington et al., 2007). In the

following analysis we employ a quantitative analysis of

topography using the evaluation of geomorphometric param-

eters (see review in Mark, 1975). Specifically, we use the

eigenvalue ratio method (see discussion in Appendix S1 and

Fig. S1) to quantify topographic roughness, organization and

gradient across the landscape (Chapman, 1952; Watson, 1966;

Woodcock, 1977; Guth, 1999; McKean & Roering, 2003). As

7915'W 7912'W 7909'W

248'S

245'S

3500 4000

Topography (m)

7915'W 7912'W 7909'W

248'S

245'S

0.0 0.2 0.4 0.6 0.8 1.0

Roughness

7915'W 7912'W 7909'W

248'S

245'S

0 1 2 3 4 5

Organization

7915'W 7912'W 7909'W

248'S

245'S

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

Gradient

(a) (b)

(c) (d)

Figure 3 The spatial relationship between the (a) topography, (b) topographic roughness, (c) topographic organization, and (d) topo-

graphic gradient for the study area shown in Fig. 1 computed using the eigenvalue ratio method. The white-filled polygons designate

the spatial extent of the tree islands, which are generally located near regions of high topographic roughness, high gradient and low

organization, suggesting a correlation between the location of the tree islands in Cajas National Park and the rough topographic features in

the landscape.




6/13

discussed above, our study is motivated by the hypothesis that

the tree islands in Cajas National Park are located in areas that

experience lower fire frequencies owing to local topographic

variations such as steep terrain. We test this hypothesis by

determining spatial patterns of geomorphometric parameters

and relating them to the locations of the tree islands. In

addition, through the use of histograms and other graphical

techniques, we demonstrate relationships among topographic

variables, elevation, and size of forest patches. A summary of

the topographic analysis methodology used to quantify various

geomorphometic parameters is provided in Appendix S1.

Because the objectives of our topographic analysis require

precise measurements, we constructed a 30-m DEM. Althoughthis resolution gives a threefold improvement in the topo-

graphic resolution over commonly available DEMs (e.g. SRTM

data), we acknowledge that it is still too coarse for a reliable

evaluation of the metre-scale topographic details that would be

required for a study of individual tree islands. The conclusions

of our study are therefore limited to topographic features with

a wavelength of about several tens of metres, and should be

considered a first-order analysis of the broad-scale topographic

characters that contribute to the existence and distribution of

the tree islands in Cajas National Park. A more detailed

geomorphometric analysis of individual tree island locales is

left for future research efforts when higher-resolution DEMs

(< 10 m) become available.

RE S ULT S

The study area, shown in Fig. 3a, covers an area of c. 90 km2

and has considerable topographic relief, with elevations

ranging from 3361 to 4406 m a.s.l., with a mean and standard

deviation of 3892 and 115 m, respectively. In general, the tree

islands (white polygons in Fig. 3) are located in close

proximity to areas of high topographic roughness (red regions

in Fig. 3b) and low topographic organization (Fig. 3c). Areasof high roughness generally correspond to high topographic

gradients (Fig. 3d). The correlation between tree island

locations and steep/rough topography is stronger for the

larger tree island areas (> 10 ha), which also tend to be located

at high elevations (at an average of 3900 m vs. about 3850 m

for the smaller tree island patches). As shown in Fig. 4, the

elevation distributions for the larger tree islands are also

skewed towards higher elevations with a smaller standard

deviation for the distribution (c. 95 m vs. a standard deviation

of > 120 m for the tree island sites covering smaller areas).

0

2

4

6

8

10

%

3000 3500 4000 4500

Elevation (metres)

0

2

4

6

8

10

%

3000 3500 4000 4500

Elevation (metres)

0

5

10

15

%

3000 3500 4000 4500

Elevation (metres)

0

5

10

15

%

3000 3500 4000 4500

Elevation (metres)

0

5

10

15

20

25

%

3000 3500 4000 4500Elevation (metres)

Random 10 ha

N: 570 mean: 3892 m N: 2010

mean: 3873 m

N: 6659 mean: 3879 m N: 4289 mean: 3855 m

N: 2785

mean: 3909 mFigure 4 Histograms of the elevation dis-

tribution for tree island sites with areal

extents of < 1 ha, 15 ha, 510 ha, and

> 10 ha. For reference, the elevation distri-

bution for 570 random sites in the study area

is also shown. Although the average elevation

of the various tree island sites is similar to the

average elevation of the random sites, there

are significant differences in the topographic

character (Figs 5 & 6).




7/13

Variations in the character of the topographic roughness

and topographic organization between the tree island sites and

a reference set of 481 sites (randomly selected within the study

area) is shown in Fig. 5, with mean and standard deviation

values listed in Table 1. Several generalizations can be made on

the basis of these distributions. The low topographic organi-

zation for the highest tree island sites (blue) is indicative of a

high degree of scatter in the topographic roughness distribu-

tion (quite similar to the case for a randomly oriented

topographic distribution) possibly reflecting the geomorphic

processes of Pleistocene-era glaciation. Moreover, within the

highest-elevation belt, grass height and coverage are typically

lower, and other life forms such as cushion plants increase in

importance. Fires not only are less likely to be ignited at highelevations, but are less likely to consume this relatively

heterogeneous vegetation.

The topographic character of the smallest size class (< 1 ha)

is closest to that of the random sites. This suggests that, given

the spatial limitation of the DEM used in this analysis (cell size

of 0.09 ha), we may not have detected terrain features that

influence the locations of these islands. Tree islands at lower

elevations (red distributions) have the most variable and

highest organization, and consistently higher roughness values

which is expected given the mechanism by which glaciers

carve landscapes. We also note that the difference between the

lower (red) and higher (blue) tree island sites increases with

forest island area in both the roughness and organization

distributions (Fig. 5). Sites with high values of both organi-

zation and roughness (e.g. lower-elevation tree islands with

spatial extent greater than 10 ha) span regions with greater

topographic variation, and the higher coherence in the surface-

normal vectors (which define the surface orientation) in the

flatter regions increases the organization values compared with

those in the rougher regions. As expected, this effect is more

pronounced for the larger tree island sites.

We applied the KolmogorovSmirnov test (KS-test) to

determine if the tree island histograms for each elevation range

and size subset differed significantly from the random sampledistribution; this test has the advantage that it is not based on

any underlying distribution of data. The D-statistic listed in

Table 1 is a measure of the difference between the randomly

selected distribution and the distribution at the listed elevation

and size subset. The confidence level that the histograms are

sampling different distributions is reflected in the significance

level probability. For reference, identical distributions would

return values of 0.0 and 0% confidence for D and the

confidence probability, respectively. The null hypothesis that

the histograms are sampling the same distribution can be

0

5

10

15

0.0 0.5 1.0

Roughness

Random

0.5 1.0 0.5 1.0 0.5 1.0 0.5 1.0

0

5

10

Organization

Random

0 4 8 0 4 8 0 4 8 0 4 8 0 4 8

(a)

(b)

Figure 5 (a) Histograms showing the elevation dependence on the topographic roughness for the tree island sites with areal extents of

< 1 ha, 15 ha, 510 ha and > 10 ha, and a reference set of 481 sites (Random). (b) Histograms showing the elevation dependence on the

topographic organization for the tree island sites with areal extents of < 1 ha, 15 ha, 510 ha and > 10 ha, and a reference set of 481 sites

(Random). The histograms are colour-coded to correspond to three elevation ranges: red (35003700 m), green (37003900 m) and blue

(39004100 m); the number of samples for each of the elevation ranges is also colour-coded. Averages, standard deviations, n-values and the

results of the KolmogorovSmirnov test applied to the distributions are listed in Table 1.




8/13

rejected at a > 99% confidence level for nearly all classes, with

the exception of the high-elevation subregion for 15 ha

patches and the lowest-elevation subregion for 510 ha tree

island patches although we note that the KS-test confidence

level for these two regions is still in excess of 75%. These results

support our a priori assumption that the geomorphological

parameters of the tree island locations are significantly

different from those of the region (as measured by a random

sampling).

The trade-off between topographic roughness and topo-

graphic organization can be graphically illustrated given the

fact that only two of the three eigenvalues are independent.

The flatness [ln(S1/S2)] (the reciprocal of roughness) can be

plotted against organization [ln(S2/S3)] to eliminate one of the

dependent variables and to provide a way to describe thepattern of vector orientations as clusters (unimodal distribu-

tion of the vectors) or girdles (multimodal distribution of the

vectors) in K-space (cf. Fig. S2 and Woodcock, 1977). The

K-space distributions computed from the flatness and organ-

ization values for Cajas National Park are shown in Fig. 6, in

which much of the information present in the histogram

distributions is distilled graphically. In general, the lower tree

island sites (black symbols) have greater organization (reflect-

ing greater coherence in the surface-normal vector distribu-

tions) but lower flatness (higher roughness) values. In contrast,

the largest tree island sites (inverted triangles) are the least

organized, with significantly greater flatness values. Again, this

reflects the fact that the larger tree island sites include large

portions of the flatter down-slope landscape at the base of

high-relief topography. We also note that lower tree island sites

lie closer to the cluster-girdle threshold defined by K = 1,

whereas the higher tree island locations have distributions

more characteristic of clusters (see Fig. S2 for an explanation

of K-space).

DIS CUS S IO N

Numerous authors have expressed concern about the ongoing

degradation of upper montane tropical ecosystems (e.g.

Millones, 1982; Hess, 1990). Polylepis forests are clearlyamong South Americas most endangered ecosystems (Ro-

moleroux, 1994; Navarro et al., 2005); however, here we argue

that, in the case of Cajas National Park, most of the forest

removal probably occurred in the distant past. As a result of a

long history of several land-use practices, especially wood

gathering and grazing (Ellenberg, 1979; Lgaard, 1992), this

ecosystem is currently restricted to sites where burning is

strongly inhibited. Extensive field observations within our

study area suggest that burning rarely, if ever, occurs within

the boulder-strewn areas at the bases of cliffs, near road cuts,

Table 1 Summary of roughness and

organization statistics for the histogram

distributions shown in Fig. 5 and the results

of the KolmogorovSmirnov test. Elevation

range n

Average 1 standard

deviation

KolmogorovSmirnov test

(D-statistic, significance level

probability [%])*

Roughness Organization Roughness Organization

Random

35003700 m 98 0.33 0.88 1.77 0.96

37003900 m 160 0.29 0.85 1.23 0.79

39004100 m 223 0.30 0.97 1.08 0.61

< 1 ha

35003700 m 121 0.36 0.62 1.90 0.71 0.21, 98.82 0.22, 99.30

37003900 m 749 0.30 0.80 1.34 0.70 0.13, 98.64 0.18, >99.99

39004100 m 851 0.31 0.77 0.93 0.57 0.14, 99.98 0.15, >99.99

15 ha

35003700 m 978 0.41 0.77 2.26 0.86 0.36, >99.99 0.36, >99.99

37003900 m 1658 0.31 0.91 1.52 0.84 0.12, 98.11 0.18, >99.99

39004100 m 2704 0.30 0.98 1.12 0.62 0.09, 95.6 0.07, 76.1

510 ha

35003700 m 908 0.38 0.65 1.75 1.02 0.25, >99.99 0.11, 79.8

37003900 m 1016 0.43 0.64 2.06 0.55 0.55, >99.99 0.53, >99.99

39004100 m 1987 0.35 0.65 1.41 0.68 0.34, >99.99 0.26, >99.99

> 10 ha

35003700 m 950 0.41 0.54 2.00 0.94 0.37, >99.99 0.20, 99.82

37003900 m 668 0.22 0.74 0.80 0.27 0.51, >99.99 0.38, >99.99

39004100 m 1109 0.24 0.72 0.73 0.36 0.37, >99.99 0.32, >99.99

*The KolmogorovSmirnov test (KS-test) results for each elevation range and area subset are

based on a comparison with the random distribution. The KS-test parameters listed are first the

D-statistic (a measure of the difference between the two distributions) and second the confidence

probability that the data sets are drawn from the different distributions. For reference, identical

distributions would return values of 0.0 and 0% for D and the confidence probability,

respectively.




9/13

or within very moist areas. Moreover, fires are relatively

unlikely to consume vegetation located above moist ravines or

depressions, where soil moisture levels are very high (Keating

et al., 2002).

Although the works cited above mention similar factors

concerning Polylepis distribution, this is, to our knowledge,

the first article to quantify the correlation between the

occurrence of Polylepis forests and terrain characteristics.

Clearly, additional studies involving extensive fieldwork will

be required to elucidate the fine-scale processes that govern

fire movement and the mortality of vegetation. Whereas

numerous models have been developed to simulate fire

behaviour in temperate forests (Keane et al., 2004), no such

modelling efforts, to our knowledge, have been applied to

high-elevation grasslands. The results of fire simulationefforts are likely to confirm the observations made in El

Cajas and elsewhere, but one should bear in mind that the

impacts of fire on vegetation vary with many factors,

including time of year, combinations of disturbances (espe-

cially grazing and burning), dominant species present, and

elevation (Verweij & Budde, 1992; Horn, 1998; Keating,

1998). One should also note that the interactions between

fires, grazing and other anthropogenic disturbances are very

complex and certainly merit further research (Renison et al.,

2006; Cierjacks et al., 2007).

The tropical Andes are well known for their species

diversity, endemism, and habitat diversity (Gentry, 1995;

Jrgensen & Leon-Yanez, 1999; Luteyn, 1999; Young et al.,

2002). Given both the high rates of land-cover conversion

and the growing interest in conservation issues, the impor-

tance of mapping and analysing tropical montane ecosystems

at the landscape scale is obvious. Most digital mapping

projects for such areas have focused on either Central

America or Mexico (e.g. Sader et al., 1989; Helmer et al.,

2000; Nagendra et al., 2004; Aguilar, 2005). A growing

number of mapping projects for the tropical Andes have

been performed at the regional to subcontinental scale (e.g.

Sierra et al., 1999), but thus far relatively few studies have

involved mapping tropical Andean vegetation at the land-

scape scale (e.g. Colby & Keating, 1998; Echavarria, 1998).

Even fewer have included GIS analyses or modelling to

explain how vegetation is influenced by disturbance or terrain

features (e.g. Keating, 1997; Vanacker et al., 2003; Kintz et al.,

2006). In part this trend stems from the scarcity of cloud-free

images for these regions, as well as from the difficulties

associated with performing the necessary fieldwork. For agrowing number of areas, however, digital data sets are

becoming available, and we recommend that upper montane

ecosystems be studied with geospatial techniques when

possible.

Here we have taken advantage of DEMs to evaluate

quantitatively the geomorphometric parameters of the land-

forms in Cajas National Park and to evaluate correlations

between landforms and the occurrence of tree islands. We

reiterate and readily acknowledge that it is an oversimplifica-

tion to claim that the distribution of tree islands is controlled

solely by the topographic character. Our working hypothesis,

however, is that certain topographic features inhibit the

movement of fire and therefore influence the current spatial

distribution of tree islands in the Cajas region. The close

proximity of the tree islands in Cajas National Park to areas of

high topographic roughness, high gradient and low organiza-

tion supports the notion that rough terrain encourages the

persistence of tree islands in this area. Forest fragments are also

highly threatened by burning in some lowland tropical

ecosystems (Gascon et al., 2000; Cochrane, 2001), but, even

in these environments, topographic features or rocky sub-

strates may afford some protection (e.g. Fordyce et al., 1997).

Our twin goals were to demonstrate the utility of

performing a quantitative topographic analysis and to evaluate

to what degree topography controls the tree island distribu-tion. Other influences are certainly important, but quantifica-

tion of their role is beyond the scope of this study. These

factors include grazing intensity, variance in precipitation,

bedrock geology of the region, hydrology, and other topo-

graphic parameters. We also acknowledge that micro-climatic

factors are very important in governing the position of

treelines. Recent work demonstrates unequivocally that the

relationships between microclimate and fire ecology are very

complex (Bader, 2007). Regardless of the relative importance

of other factors that we did not measure, this article

0

2

4

6

Flatnessln(S1/S2)

0 2 4 6

Organization ln(S2/S3)

Clusters

Girdles

Streng

th

Figure 6 The location in K-space [a plot of organization (ln(S2/

S3) vs. flatness (ln(S1/S2)] for the tree island locations in three

elevation ranges, 35003700 m (black), 37003900 m (grey),

39004100 m (white), and for areal extents of < 1 ha, 15 ha, 5

10 ha, and > 10 ha. Also shown are the locations for the random

sites in the study area. As discussed in Appendix S1, the variation

in the K-space location for the various subregions provides a way

to quantify the topographic differences among the various regions.

The n-values for the distributions are listed in Table 1.




10/13

demonstrates that the position of forest patches in El Cajas is

indeed consistent with what is currently understood about fire

behaviour. These results should motivate a more detailed study

of the small-scale (< 10 m resolution) topography of individ-

ual tree islands as such data become available (most likely

requiring 1-m resolution LiDAR data).

CO NCLUS IO NS

In this article we have tested the hypothesis that the spatial

distribution of high-elevation tropical forest patches in Cajas

National Park is controlled by the landscape-scale topographic

character by evaluating the correlation between the tree island

distribution and a number of geomorphometric parameters

including organization, roughness and spectral power. Within

the assumptions and resolution limitations of our study, we

draw the following conclusions regarding the correlation

between features in the topography and the location of the

tree islands:

(1) A quantitative topographic analysis is a valuable tool for

understanding the distribution of land cover, particularly inregions where insular mountain ranges dominate the topo-

graphic distribution.

(2) Whereas climatic factors may have determined the original

distribution ofPolylepis-dominated forests, it appears that fires

exert a strong influence on the present distribution of these

forests. Our quantitative analyses demonstrate that areas at the

base of cliffs or above moist valleys are among the few places

where these ecosystems persist in south-western Ecuador.

(3) The geomorphometric analysis of the topography indicates

that the tree islands of Cajas National Park are found in close

proximity to rough topography with strong topographic

gradients. Overall, we find that the topographic roughness

exerts the primary control on the topographic character of the

landscape in the immediate area surrounding the tree islands.

(4) Climate, terrain, and disturbances all exert strong influ-

ences on the spatial position and areal extent of upper

montane tropical forests. The interrelationships among these

factors merit far more attention than they have previously been

accorded. As technology improves, scientists ability to relate

the position of plant communities to numerous variables will

increase further. In the future, extensive field data could be

coupled with model simulations to enhance the understanding

of tropical forest dynamics.

ACK NO W LE DGE M E NT S

Research for this article was funded by the Fulbright

Commission of Ecuador, the National Geographic Society,

Indiana University, and NSF Grant EAR-9422423 awarded to

B. Hansen, B. Leon, D. Rodbell, G. Seltzer, and K. R. Young.

The Instituto Geografico Militar in Quito provided topo-

graphic maps and aerial photographs, and LAND INFO

Worldwide Mapping, LLC created the DEM utilized in this

study. We are grateful to INEFAN for providing a permit to

conduct research within the park, to the herbarium staff at the

Catholic University in Quito for providing facilities and

logistical support, and to Stephanie Dickenson and Tom Evans

for providing statistical consulting. The program GMT (Wessel

& Smith, 1991) was used for a large part of the analysis as well

as for the figures in this paper.

RE FE RE NCE S

Aguilar, A. (2005) Remote sensing of forest regeneration in

highland tropical forests. GIScience and Remote Sensing, 42,

6679.

Bader, M.Y. (2007) Tropical alpine treelines: how ecological

processes control vegetation patterning and dynamics. PhD

Thesis, Wageningen University, Wageningen, The Nether-

lands.

Blandn, L.C. (1989) Analisis y estudios climatolo gicos en el

Ecuador. Biblioteca Ecuador, Quito.

Borrero, V.A.L. (1989) Paisaje rural en el Azuay. Biblioteca de

geografa ecuatoriana, Vol. 5. Banco Central del Ecuador,

Quito.

Bowman, D.M.J.S. (2000) Australian rainforests: islands of green in a land of fire. Cambridge University Press, Cam-

bridge.

Brown, J.H. & Gibson, A.C. (1983) Biogeography. Mosby Co.,

St Louis.

Bruhns, K.O. (1994) Ancient South America. Cambridge Uni-

versity Press, Cambridge.

Bruhns, K.O., Burton, J.H. & Miller, G.R. (1990) Excavations

at Pirincay in the Paute Valley of southern Ecuador, 1985

1988. Antiquity, 64, 221233.

Chapman, C.A. (1952) A new quantitative method of

topographic analysis. American Journal of Science, 250, 428

452.

Chorley, R.J., Malm, D.E.C. & Pogorzelski, H.A. (1957) A new

standard for measuring drainage basin shape. American

Journal of Science, 255, 138141.

Cierjacks, A., Wesche, K. & Hensen, I. (2007) Potential lateral

expansion of Polylepis fragments in central Ecuador. Forest

Ecology and Management, 242, 477486.

Clapperton, C.M. (1993) The glaciation of the Andes. Qua-

ternary Science Reviews, 2, 83155.

Coblentz, D. & Riitters, K.H. (2004) Topographic controls on

the regional-scale biodiversity of the southwestern USA.

Journal of Biogeography, 31, 11251138.

Cochrane, M.A. (2001) Synergistic interactions between habi-

tat fragmentation and fire in evergreen tropical forests.Conservation Biology, 15, 15151521.

Colby, J.D. & Keating, P.L. (1998) Land cover classification

using landsat TM imagery in the tropical highlands: the

influence of anisotropic reflectance. International Journal of

Remote Sensing, 19, 14791500.

Collier, D. & Murra, J.V. (1943) Survey and excavations in

Southern Ecuador. Anthropological Series 35, Field Museum

of Natural History, Chicago.

Cuatrecasas, J. (1968) Paramo vegetation and its life forms.

Geo-ecology of the mountainous regions of the tropical




11/13

Americas (ed. by C. Troll), pp. 163186. Colloquium Geo-

graphicum 9, Geographisches Institut, Bonn.

Echavarria, F.R. (1998) Monitoring forests in the Andes

using remote sensing: an example from southern Ecuador.

Natures geography: new lessons for conservation in

developing countries (ed. by K.S. Zimmerer and K.R.

Young), pp. 100120. University of Wisconsin Press,

Madison.

Ellenberg, H. (1979) Mans influence on tropical mountain

ecosystems in South America. Journal of Ecology, 67, 401

416.

Enrico, L., Funes, G. & Cabido, M. (2004) Regeneration of

Polylepis australis Bitt. in the mountains of central

Argentina. Forest Ecology and Management, 190, 301

309.

Fehse, J., Hofstede, R., Aguirre, N., Paladines, C., Kooijman, A.

& Sevink, J. (2002) High altitude tropical secondary forests:

a competitive carbon sink? Forest Ecology and Management,

163, 925.

Fensham, R.J. (1995) Floristics and environmental relations of

inland dry rainforests in north Queensland, Australia. Journal of Biogeography, 22, 10471063.

Fjeldsa, J. (1992) Biogeography of the birds of the Polylepis

woodlands of the Andes. Paramo: an Andean ecosystem

under human influence (ed. by H. Balslev and J.L. Luteyn),

pp. 3144. Academic Press, London.

Fordyce, I.R., Eamus, D., Duff, G.A. & Williams, R.J. (1997)

The role of seedling age and size in the recovery of Allo-

syncarpia ternata following fire. Australian Journal of Ecol-

ogy, 22, 262269.

Forman, R.T.T. & Godron, M. (1986) Landscape ecology. John

Wiley, New York.

Fundacion Natura (1992) Acciones de desarrollo en zonas de

influencia de areas protegidas. Fundacion Natura, Quito,

Ecuador.

Gade, D.W. (1999) Nature and culture in the Andes. University

of Wisconsin Press, Madison, WI.

Gascon, C., Williamson, G.B. & de Fonseca, G.A.B. (2000)

Receding edges and vanishing fragments. Science, 288, 1356

1358.

Gentry, A.H. (1995) Patterns of diversity and floristic com-

position in neotropical montane forests. Biodiversity and

conservation of neotropical montane forests (ed. by S.P.

Churchill, H. Balslev, E. Forero and J.L. Luteyn), pp. 103

126. New York Botanical Garden, The Bronx, NY.

Guth, P.L. (1999) Quantifying and visualizing terrain fabricfromdigitalelevation models. Geocomputation 99: Proceedings

of the 4th International Conference on GeoComputation (ed. by

J. Diaz, R. Tynes, D. Caldwell and J. Ehlen). Available at:

http://www.geovista.psu.edu/geocomp/geocomp99/index.htm

(accessed 27 May 2008).

Hansen, B.C.S., Rodbell, D.T., Seltzer, G.O., Leon, B., Young,

K.R. & Abbott, M. (2003) Late-glacial and Holocene vege-

tational history from two sites in the western Cordillera of

southwestern Ecuador. Paleogeography, Paleoclimatology,

Paleoecology, 194, 79108.

Harden, C. (2007) Two soil toposequences in a tropical glacial

trough, southern Ecuador. Zeitschift fur Geomorphologie, 51,

139152.

Hastenrath, S. (1981) The glaciation of the Ecuadorian Andes.

A.A. Balkema, Rotterdam.

Helmer, E.H., Brown, S. & Cohen, W.B. (2000) Mapping

montane tropical forest successional stage and land use with

multidate Landsat imagery. International Journal of Remote

Sensing, 21, 21632183.

Hensen, I. (2002) Impacts of anthropogenic activity on the

vegetation of Polylepis woodlands in the region of Coc-

habamba/Bolivia. Ecotropica, 8, 182203.

Hess, C.G. (1990) Moving up - moving down: agro-pastoral

land-use patterns in the Ecuadorian paramos. Mountain

Research and Development, 10, 333342.

Horn, S.P. (1998) Fire management and natural landscapes in

the Chirripo Paramo, Chirripo National Park, Costa Rica.

Natures geography: new lessons for conservation in developing

countries (ed. by K.S. Zimmerer and K.R. Young), pp. 125

146. University of Wisconsin Press, Madison, WI.

Ibisch, P. (1993) Estudio de la vegetacion como una contributiona la caracterizacion de la provincia de Arque (Bolivia). Cu-

aderno Cientfico No. 1 PROSANA, Cochabamba, Bolivia.

Jensen, J.R. (2005) Digital image processing: a remote sensing

perspective, 3rd edn. Prentice Hall, Englewood Cliffs, NJ.

Jimenez de Espada, M. (1965[1582]) Relaciones geograficas de

Indias. Biblioteca de autores espanoles, Vol. 183. Ediciones

Atlas, Madrid.

Jrgensen, P.M. & Leon-Yanez, S., eds (1999) Catalogue of

the vascular plants of Ecuador. Monographs in Systematic

Botany from the Missouri Botanical Garden, Vol. 75, pp. 1

1182. The Missouri Botanical Garden Press, St Louis, MO.

Keane, R.E., Cary, G.J., Davies, I.D., Flannigan, M.D., Gardner,

R.H., Lavorel, S., Lenihan, J.M., Li, C. & Rupp, T.S. (2004) A

classification of landscape fire succession models: spatial

simulations of fire and vegetation dynamics. Ecological

Modelling, 179, 327.

Keating, P.L. (1997) Mapping vegetation and anthropogenic

disturbances in southern Ecuador with remote sensing

techniques: implications for park management. Yearbook,

Conference of Latin Americanist Geographers, 23, 7790.

Keating, P.L. (1998) Effects of anthropogenic disturbances on

paramo vegetation in Podocarpus National Park, Ecuador.

Physical Geography, 19, 221238.

Keating, P.L. (1999) Changes in paramo vegetation along an

elevation gradient in southern Ecuador. Journal of the TorreyBotanical Society, 126, 159172.

Keating, P.L. (2007) Fire ecology and conservation in the high

tropical Andes: observations from northern Ecuador. Jour-

nal of Latin American Geography, 6, 4362.

Keating, P.L., Young, K.R. & Leon, B. (2002) Variation in high

Andean vegetation at a site in southwestern Ecuador.

Pennsylvania Geographer, 40, 1535.

Kessler, M. (1995) Polylepis-Walder Boliviens: Taxa, Okologie,

Verbreitung und Geschichte. Dissertationes Botanicas 246.

Science Publishers, Stuttgart, Germany.




12/13

Kintz, D.B., Young, K.R. & Crews-Meyer, K.A. (2006) Impli-

cations of land-use/land cover change in the buffer zone of a

national park in the tropical Andes. Environmental Man-

agement, 38, 238252.

Lgaard, S. (1992) Influence of fire in the grass paramo veg-

etation of Ecuador. Paramo: an Andean ecosystem under

human influence (ed. by H. Balslev and J.L. Luteyn),

pp. 151170. Academic Press, New York.

Lauer, W. (1981) Ecoclimatological conditions of the paramo

belt in the tropical high mountains. Moutain Research and

Development, 1, 209221.

Luteyn, J.L. (1999) Paramos: a checklist of plant diversity, geo-

graphical distribution, and botanical literature. Memoirs of the

New York Botanical Garden, Vol. 84. New York Botanical

Garden Press, The Bronx, NY.

Madsen, J.E. & llgaard, B. (1994) Floristic composition,

structure, and dynamics of an upper montane rain forest in

southern Ecuador. Nordic Journal of Botany, 14, 403423.

Mark, D.M. (1975) Geomorphometric parameters: a review

and evaluation. Geografiska Annaler, Series A, Physical

Geography, 57, 165177.Mausel, P., Wu, Y., Li, Y., Moran, E.F. & Brondizio, E.S. (1993)

Spectral identification of successional stages following

deforestation in the Amazon. Geocarto International, 4, 6171.

McKean, J. & Roering, J. (2003) Objective landslide detection

and surface morphology mapping using high-resolution

airborne laser altimetry. Geomorphology, 57, 331351.

Millones, O.J. (1982) Patterns of land use and associated

environmental problems of the Central Andes: an integrated

summary. Mountain Research and Development, 2, 4961.

Ministerio de Ambiente (2007) Parque Nacional El Cajas.

Available at: http://www.ambiente.gov.ec (accessed 15

October 2007).

Morris, A. (1985) Forestry and land-use conflicts in Cuenca,

Ecuador. Mountain Research and Development, 5, 183

196.

Nagendra, H., Tucker, C., Carlson, L., Southworth, J., Kar-

macharya, M. & Karra, B. (2004) Monitoring parks through

remote sensing: studies in Nepal and Honduras. Environ-

mental Management, 34, 748760.

Navarro, G., Monina, J.A. & De La Barra, N. (2005) Classifi-

cation of the high-Andean Polylepis forests in Bolivia. Plant

Ecology, 176, 113130.

Renison, D., Hensen, I., Suarez, R. & Cingolani, A.M. (2006)

Cover and growth habit of Polylepis woodlands and shrub-

lands in the mountains of central Argentina: human orenvironmental influence? Journal of Biogeography, 33, 876

887.

Riitters, K.H., Wickham, J.D., Vogelmann, J.E. & Jones, K.B.

(2000) National land-cover pattern data. Ecology, 81, 604.

Rodbell, D.T., Bagnato, S., Nebolini, J.C., Seltzer, G.O. &

Abbott, M.B. (2002) A late glacial-Holocene tephrochro-

nology for glacial lakes in southern Ecuador. Quaternary

Research, 57, 343354.

Romoleroux, K. (1994) The family Rosaceae in Ecuador. PhD

Thesis, University of Aarhus, Aarhus, Denmark.

Sader, S.A., Waide, R.B., Lawrence, W.T. & Joyce, A.T. (1989)

Tropical forest biomass and successional age class relation-

ships to a vegetation index derived from Landsat TM data.

Remote Sensing of Environment, 28, 143156.

Sappington, J.M., Longshore, K.M. & Thompson, D.B. (2007)

Quantifying landscape ruggedness for animal habitat anal-

ysis: a case study using bighorn sheep in the Mojave Desert.

Journal of Wildlife Management, 71, 14191426.

Sarmiento, F.O. (1995) Restoration of equatorial Andes: the

challenge for conservation of Trop-Andean Landscapes in

Ecuador. Biodiversity and conservation of Neotropical mon-

tane forests (ed. by S. Churchill, H. Balslev, E. Forero and J.L.

Luteyn), pp. 637665. The New York Botanical Garden, The

Bronx, NY.

Sierra, R., Ceron, C., Palacios, W. & Valencia, R. (1999) Mapa

de vegetacion del Ecuador continental. 1: 1,000,000. Proyecto

INEFAN/GEF-BIRF. Wildlife Conservation Society and

Ecociencia, Quito.

Simpson, B.B. (1979) A revision of the genus Polylepis (Ros-

aceae: Sanguisorbeae). Smithsonian Contributions to Botany,

43, 162.Simpson, B.B. (1986) Speciation and specialization of Polylepis

in the Andes. High altitude tropical biogeography (ed. by F.

Vuillemier and M. Monasterio), pp. 304316. Oxford Uni-

versity Press and the American Museum of Natural History,

New York.

Sklenar, P. & Jrgensen, P.M. (1999) Distribution patterns of

paramo plants in Ecuador. Journal of Biogeography, 26, 681

691.

Sklenar, P. & Ramsay, P.M. (2001) Diversity of zonal paramo

plant communities in Ecuador. Diversity and Distributions,

7, 113124.

Stern, M.J. (1995) An inter-Andean forest relict: vegetation

change on Pasochoa Volcano, Ecuador. Mountain Research

and Development, 15, 339348.

Troll, C. (1968) The cordilleras of the tropical Americas: as-

pects of climatic, phytogeographical, and agrarian ecology.

Geoecology of mountainous regions of the tropical Americas

(ed. by C. Troll), pp. 1555. Colloquium Geographicum 9,

Geographisches Institut, Bonn.

Troll, C. (1973) The upper timberlines in different climatic

zones. Arctic and Alpine Research, 5, A3A18.

Turner, M.G., Gardner, R.H. & ONeill, R.V. (2001) Landscape

ecology in theory and practice: pattern and process. Springer,

New York.

Valencia, R. & Jrgensen, P.M. (1992) Composition andstructure of a humid montane forest on the Pasochoa Vol-

cano, Ecuador. Nordic Journal of Botany, 12, 239247.

Vanacker, V., Vanderschaeghe, M., Govers, G., Willems, E.,

Poesen, J., Deckers, J. & De Bievre, B. (2003) Linking

hydrological, infinite slope stability and land-use change

models through GIS for assessing the impact of deforesta-

tion on slope stability in high Andean watersheds. Geo-

morphology, 52, 299315.

Verweij, P.A. & Budde, P.E. (1992) Burning and grazing

gradients in paramo vegetation: initial ordination analyses.




13/13

Paramo: an Andean ecosystem under human influence (ed. by

H. Balslev and J.L. Luteyn), pp. 177196. Academic Press,

London.

Watson, G.S. (1966) The statistics of orientation data. Journal

of Geology, 74, 786797.

Wessel, P. & Smith, W.H.F. (1991) Free software helps map

and display data. EOS Transactions of the American Geo-

physical Union, 72, 441.

White, S. & Maldonado, F. (1991) The use and conservation of

natural resources in the Andes of southern Ecuador.

Mountain Research and Development, 11, 3755.

Woodcock, N.H. (1977) Specification of fabric shapes using an

eigenvalue method. Geological Society of America Bulletin,

88, 12311236.

Young, K.R.(1994) Roads and the environmental degradation of

tropical montane forests. Conservation Biology, 8, 972976.

Young, K.R. (1998a) Composition and structure of a timber-

line forest in north-central Peru. Forest biodiversity in North,

Central and South America, and the Caribbean: research and

monitoring(ed. by F. Dallmeier and J.A. Comisky), pp. 595

613. Parthenon Publishing, Pearl River, NY.Young, K.R. (1998b) Deforestation in landscapes with humid

forests in the Central Andes: patterns and processes. Nat-

ures geography: new lessons for conservation in developing

countries (ed. by K.S. Zimmerer and K.R. Young), pp. 7599.

University of Wisconsin Press, Madison, WI.

Young, K.R. & Keating, P.L. (2001) Remnant forests of Volcan

Cotacachi, northern Ecuador. Arctic, Antarctic and Alpine

Research, 33, 165172.

Young, K.R. & Zimmerer, K.S. (1998) Conclusion: biological

conservation in developing countries. Natures geography:

new lessons for conservation in developing countries (ed. by

K.S. Zimmerer and K.R. Young), pp. 327344. University of

Wisconsin Press, Madison, WI.

Young, K.R., Ulloa Ulloa, C., Luteyn, J.L. & Knapp, S. (2002)

Plant evolution and endemism in Andean South America:

an introduction. The Botanical Review, 68, 421.

S UP P O RT ING INFO RM AT IO N

Additional Supporting Information may be found in the

online version of this article:

Appendix S1 Topographic analysis and the eigenvalue ratio

method.

Figure S1 Example surface-normal vectors illustrating topo-

graphic surface roughness.

Figure S2 Illustration ofK-space (topographic flatness plotted

vs. the topographic organization) and the location of several

example topographic distributions.

Please note: Wiley-Blackwell is not responsible for the content

or functionality of any supporting materials supplied by the

authors. Any queries (other than missing material) should be

directed to the corresponding author for the article.

B I O S K E T C H E S

David Coblentz is a solid Earth geophysicist with a joint

appointment at Los Alamos National Laboratory and the

University of New Mexico who, through speculative geody-

namics research, explores the common ground between

geodynamics and geoecology.

Phil Keating has spent 17 years investigating the influences

of anthropogenic disturbances on vegetation in the Ecuado-

rian Andes. He often combines field-based ecological studies

with remote sensing and GIS to determine how different fire

regimes modify floristic composition and landscape structure

in both paramo and timberline forest communities. DrKeating is currently a research associate at Indiana University.

Editor: Jorge Crisci



Documents

Topographic Controls on the Distribution of Tree Islands in the High Andes of South-western Ecuador