Environmental filters (e.g. climate, geomorphology and soils) and dispersal filters are key determinants of species distributions of Amazonian tree communities. However, a comprehensive analysis of the role of environmental and dispersal filters is needed to understand the ecological and evolutionary processes that drive phylogenetic and taxonomic turnover of Amazonian tree communities. We compare measures of taxonomic and phylogenetic beta diversity in 40 one-hectare plots to test the relative importance of climate, soils, geology, geomorphology, pure spatial variables and the spatial variation of environmental drivers of phylogenetic and taxonomic turnover in Ecuadorian Amazon tree communities. We found low phylogenetic and high taxonomic turnover with respect to environmental and dispersal filters. In addition, our results suggest that climate is a significantly better predictor of phylogenetic turnover and species turnover than geomorphology and soils at all spatial scales. The influence of climate as a predictor of phylogenetic turnover was stronger at broader spatial scales (50 km²) whereas geomorphology and soils appear to be better predictors of taxonomic turnover at mid (5 km²) and fine spatial scales (0.5 km²) but a weak predictor of phylogenetic turnover at broad spatial scales. We also found that the combined effect of geomorphology and soils was significantly higher for species turnover at all spatial scales but not for phylogenetic turnover at large spatial scales. Geographic distances as proxy of dispersal limitation was a better predictor of phylogenetic turnover at distances of 50<500 km. Our findings suggest that climatic variation at local and regional scales can better predict phylogenetic and taxonomic turnover than geomorphology and soils.

The 40 one hectare plot work used in this study are part of the Amazon Tree diversity Network and have been established by the lead author and co-authors in the past 20 years representing an unprecedented data set. In the past four years, we established 15 one-hectare plots in both terra firme and white sand forests in the Ecuadorian Amazon. We established 8 plots on alluvial terraces of the Aguarico River towards the north of the Ecuadorian Amazon; the age of these units ranges from Pliocene to Pleistocene origin (http://www.geoinvestigacion.gob..ec, Laraque et al.2009) The landscape is mostly characterized by large areas that correspond to Pleistocene alluvial terraces that occasionally suffer flooding events. These geomorphological units are interrupted only by high terraces with flat surfaces that have not suffered erosion of their surfaces (Saunders 2008; Wesselingh et al. 2006). Two additional plots were established in old alluvial terraces of Napo River, these units as well as the units located in the Aguarico River are high terraces that presumably constitute old flood plains of the previously mentioned rivers (Saunders 2008). Ten plots were established in the Pastaza megafan which is a massive alluvial deposit located in the southwestern Ecuadorian Amazon, evidence suggests that the modern megafan complex dates from the Pliocene-Pleistocene and recent alluvial processes have occurred between the last 180 000-30 000 yrs. (Rasanen et al. 1995; Bernal et al. 2011). Five plots were established in areas belonging to plateaus originated during the Cretaceous period; these geomorphological units are located in the lowest part of Cordillera del Condor below 500 m.

The remaining 25 one hectare plots were established in the Yasuní national Park and the Tigre-Corrientes watershed. The landscape in both areas is characterized by the predominance of geomorphological units such as highly dissected hills occasionally interrupted by valleys (Pitman 2000). The landscape is dominated by Curaray and Chambira formations from Miocene and Mio-Pliocene origin respectively and soils are characterized by higher nutrients content (Pitman et al. 2008).

Geomorphological variables and proximity analysis

Four geographic variables (hierarchical slope position, slope, dem and landsat) were used in the analysis describing the geomorphology and land cover features in the vicinity of the forest plots. Digital terrain elevation data for the Ecuadorian Amazon was obtained from the Shuttle Radar Topography Mission (SRTM) distributed by the USGS through the Earth Explorer platform (https://earthexplorer.usgs.gov/). The SRTM dataset has worldwide coverage of void filled elevation data at a resolution of 1 arc-second (30 meters). Topographic slope in degrees was calculated from the elevation data using the Spatial Analyst extension in ArcGIS 10.3 software from ESRI (Environmental Systems Resource Institute).

Hierarchical Slope Position identifies topographic exposure (ridge, slope, valley bottom, etc) by applying moving windows with increasing radii to a digital elevation model (DEM) (Murphy et al. 2010). The exposure is a ridge if the elevation of the center cell in the window is higher than the average of the cells in the window. The opposite case corresponds to a valley bottom or toe slope. Hierarchical integration is done by starting with exposure values for the largest (user defined) window and adding values from smaller windows if their absolute standardized values exceed the values of the larger scale map. The variable was calculated using the Geomorphometry and Gradient Metrics (version 2.0) for ArcGIS (Evans et al. 2014) using windows of radii between 2 and 10 pixels with increments of two pixels.

Slope position was measured by subtracting the Slope Position average neighbor values from a focal value. Positive values indicate that the central point is located higher than its average surroundings, while negative indicates a position lower than the average. The range of the metric depends not only on differences but also on the defined neighborhood. This metric is also referred to as Topographic Position Index (TPI) (Guisan et al. 1999).

Digital elevation models (DEM) measures the bare-earth surface based on raster grids of the elevation between two or more points. We obtained data for Ecuadorian Amazon from the SRTM 90-meter resolution Digital Elevation Model developed by the NASA. We concatenated the different mosaics of topography in ArcMap 10.5.1 using the Spatial Statistics tool.

Land cover information was obtained from a mosaic of Landsat images for the period 2010-2014 that had been created for the Ministry of the Environment of Ecuador. The mosaic was was created using the approach described in Hansen et al. (2013) that includes (i) image resampling, (ii) conversion of raw digital values (DN) to top of atmosphere (TOA) reflectance, (iii) cloud/shadow/water screening and quality assessment (QA), and (iv) image normalization. A principal components analysis was performed using the three RGB bands of the mosaic and the first component, that explained 96.4% of the variance, was used for further analysis.

For each plot, a set of circular buffers with areas of 0.5 km², 5 km² and 50-100 km². Descriptive statistics were calculated for each variable at each scale defined by the buffer areas using zonal statistics tools in ArcGIS.

Climatic variables

In order to assess the role of climatic variables in the patterns of taxonomic and phylogenetic turnover we used 19 climatic variables from Bioclim at 30 seconds of resolution as an initial set of variables.