Reconstructing Connectivity Patterns Within Pantepm Under LGM Climate

For CEDM calculation, MaxEnt 3.2.1 (Phillips et al. 2006; http://www.cs.princeton. edu/~shapire/maxent) was used to assess potential distributions of vegetation zones during LGM. Maxent is a machine-learning algorithm following the principles of maximum entropy (Jaynes 1957). In order to study responses in terms of potential distributions of vegetation zones in the Guyana Highlands, especially the tepuis under different climate scenarios, distribution data points are necessary for CEDM calculation. For this, we generated 10,000 randomly distributed points spanning from latitude 0.3° S to 8.7° N and longitude 68.7° W to 57.6° W with DIVA-GIS 5.4 (Hijmans et al. 2002; This spatial coverage was deliberately made larger than those occupied by the Pantepui region to ensure that training points later used to generate CEDMs under current-day climate encompassed the full extension of the climate landscape. Altitudinal properties at each of the 10,000 points were extracted and divided into two altitudinal range classes, termed 'upland' and 'tepui' (see McDiarmid and Donnelly 2005; MacCulloch et al. 2007): i.e. 804 points at 800-1,500 m above sea level and 65 points at >1,500 m above sea level. For a better understanding of the discrimination between these two terms see Fig. 1b.

Information on current climate was obtained from the WorldClim database, version 1.4, which is based on weather conditions recorded between 1950 and 2000 with grid cell resolution 2.5 min (Hijmans et al. 2005; It was created by interpolation using a thin-plate smoothing spline of observed climate at weather stations, with latitude, longitude and elevation as independent variables (Hutchinson 1995, 2004).

For paleoclimate, General Circulation Model (GCM) simulations from the Community Climate System Model (CCSM) assuming a temperature decrease of approximately 4 °C for the target region were provided by R. J. Hijmans (http://; Kiehl and Gent 2004). Original GCM data were downloaded from the PIMP2 website ( with spatial resolution of roughly 300 x 300 km. Surfaces were created at 0.04° spatial resolution via the following procedure: first, the difference between the GCM output for LGM and current-day conditions was calculated. These were then interpolated to the 0.04° resolution grid using the spline function in ESRI ArcInfo with the tension option. Eventually, the interpolated difference was added to the high-resolution current-day climate data set from WorldClim and LGM bioclimatic coverage using DIVA-GIS. This procedure has the dual advantages of producing data at a resolution relevant to the spatial scale of analysis and of calibrating the simulated climate change data to the actual observed climate data.

The layers for each climate scenario included the minimum and maximum temperatures and the mean precipitation per month (= 36 climatic parameters) derived from the WorldClim 1.4 interpolation model. Based on these, seven bioclimatic parameters for each climate scenario were generated with DIVA-GIS: 'annual mean temperature', 'temperature seasonality' (standard deviation of monthly mean temperature x 100), 'mean temperature of the warmest quarter', 'mean temperature of the coldest quarter', 'annual precipitation', 'precipitation of the wettest quarter' and 'precipitation of the driest quarter'. These parameters reflect the availability and range of thermal energy and humidity and are suitable for CEDM projections between different climate scenarios (e.g. Carnaval and Moritz 2008). Bioclimatic parameters were exported from DIVA-GIS as *.grd/*.gri files.

CEDMs were calculated for both altitudinal range classes separately and results are shown in Fig. 4. Maxent allows for model testing by calculation of the Area Under the Curve (AUC), referring to the ROC (Receiver Operation Characteristic) curve using e.g. 25% of the records as test points and the remaining ones for training (Hanley and McNeil 1982; Phillips et al. 2006). We acknowledge that currently there has been some doubt raised on the reliability of AUC (Lobo et al. 2008). But for a lack of a better method and because AUC has been recommended and applied for ecological applications by other authors (e.g. Pearce and Ferrier 2000; Elith et al. 2006; Phillips et al. 2006), we hereby continue using it.

AUC values range from 0.5 (i.e. random) for models with no predictive ability to 1.0 for models giving perfect predictions. According to the classification of Swets (1988), AUC values >0.9 describe 'very good', >0.8 'good' and >0.7 'useful' discrimination ability. We received 'very good' AUC values in the models for 'upland' (AUC, .. = 0.972; AUC, t = 0.969) and 'tepui' (AUC, .. = 0.998;

r v training test ' r v training


Visual comparisons of our CEDM for current conditions and elevation, within the study region confirmed high overlap (Figs. 2 and 4a). CEDMs of 'upland' and 'tepui' under LGM conditions revealed that the respective areas were apparently much larger supporting the Cool Climate hypothesis of Rull (2005a). Temporal connections between the 'tepuis' during glacial and interglacial oscillations were present and may have allowed exchange of biota. According to our models, at least, the whole upland and the Eastern massifs, such as the Ayuan, Los Testigos, Chimanta, Aprada the Eastern Tepui Chain the Cerro Yavi and Yutaje massifs (Fig. 2) were connected, whereby the remaining ones exhibited no or limited contact (Fig. 4b). These findings largely coincide with those of Rull and Nogue (2007), who performed a more simple assessment of connectivity patterns assuming a adia-batic temperature gradient and cooling of about 4.1°C.

A comparison of similarities between herpetofaunal assemblages inhabiting different massifs (Fig. 3.) showed that faunas at Cerro Yavi and Yutaje are mostly alike, followed by a clade summarizing Eastern massifs (Los Testigos, Aprada, Chimanta, Auyan and Cerro Guiaquinima). Duida-Marahuaka, Jaua, Neblina-Aracamuni and the Eastern Tepui Chain are inhabited by more different assemblages. Therefore, we conclude that the connectivity patterns suggested by our palaeoclimate CEDM and hence the Cool Climate hypothesis may explain most of the similarity among herpetofaunal assemblages observed herein, but not entirely. The relatively high similarity between Cerro Guiaquinima and the other Eastern massifs (e.g. Ayuan massif) cannot be well explained with our CEDM results showing

Ancient Climate Patterns
Fig. 4 Potential distributions of 'upland' (grey) and 'tepui' climate envelopes (black) under current (a) and Last Glacial Maximum (CCSM) conditions, 21,000 years BP (b)

no connection between them. Also, the distinctness of the herpetofauna of the Eastern Tepui Chain, which may have been connected to the other Eastern massifs under LGM conditions, is remarkable.

McDiarmid and Donnelly (2005) argued that the high level of endemic amphibian and reptile species may cast some doubt on the Cool Climate hypothesis. Our results confirm that faunal exchange between most 'tepuis' was restricted, but possible during the LGM within the Eastern massifs at least. This geographic pattern is also reflected when comparing herpetofaunal assemblages, although the high degree of endemism causes remarkable differences even between geographically close massifs (Fig. 3), e.g. the Ayuan, Chimanta and Los Testigos massifs located within a distance of less than 100 km (Fig. 2).

We expect that the extent of faunal exchange during glacial periods depended on the establishment abilities of the invaders in existing faunas. Myers and Donnelly (2001) found that some endemic species have phenotypi-cally and ecologically sibling counterparts on other tepuis. MacCulloch et al. (2007) made similar observations regarding Mt. Roraima and adjacent tepuis. This appears to be especially applicable to frogs in the genera Anomaloglossus, Oreophrynella and Pristimantis as well as in snakes in the genus Thamnodynastes. These species were suggested to comprise monophyletic highland groups rather than separate invasions from lowlands (Myers and Donnelly 2001) and, considering the geological age of the Pantepui region allowing speciation, many of them may represent sister species. These taxa occupy similar ecological niches (Myers and Donnelly 2001) which would hamper the establishment of one species at tepuis inhabited by a similar species, so much so even if 'tepuis', were connected during the glacial periods, faunal exchange may have been limited, explaining - at least in parts - the distinctness of 'tepui' herpetofaunas.

Was this article helpful?

0 0

Post a comment