• Keine Ergebnisse gefunden

IV. Explaining and predicting the distribution of the mountain forest types

IV.2.2 Environmental factors

The environmental factors were selected based on their possible potential to explain the com-position and patterns of the mountain forest types in the study area. Furthermore availability of data and spatial coverage had to be taken into account.

Some of the environmental factors just represented measured or observed conditions at plot level while others were raster datasets of the entire protected area compiled from thematic maps, remote sensing data and meteorological data models by applying GIS-tools.

Plot-specific environmental factors

Along the diagonal line of each plot three soil samples of the mineral soil were taken in a depth of 15-20 cm. These samples were mixed thoroughly and mineral soil pH was analyzed with a calibrated handheld PCE-PH 20S in a 1:2.5 soil-water ratio suspension. Evidence of past disturbances caused by agriculture, hurricanes or fires was recorded per plot. The dis-tances from each plot to the closest river digitized on the topographic maps (1:50,000) were calculated. Field estimates on total vegetation cover, herb cover and no vegetation cover were also introduced.

FIGURE IV.1. Study area and location of the 137 sampled plots (Source: Plots and trails from GPS field work; Boundaries of ABNP from Law 64-00; Villages, municipalities, roads from topographic maps 1:50,000, ICM 1983/1984 and field work; trails digitized on aerial photograph mosaic)

Protected area-wide environmental factors

Of the area-wide environmental factors one was a spectral index derived from remote sensing data, seven were related to topography, six to climate, one to insolation and one to soil humid-ity.

Spectral vegetation indices like the Normalized Difference Vegetation Index (NDVI) (ROUSE et al. 1974) have been used in numerous ecological studies, many of them reviewed by KERR &OSTROVSKY (2003), PETTORELLI et al. (2005) and GLENN et al. (2008). Most vege-tation indices make use of the fact that plant pigments absorb solar radiation in the visible electromagnetic spectrum used as a source of energy in the process of photosynthesis. In the near-infrared spectral region incoming energy is scattered by the leaf structure (JENSEN 2007).

The NDVI is based on the ratio of the reflection of solar radiation in the visible red (RED) and near-infrared (NIR) sections of the spectrum. The equation is: NDVI=(NIR-RED)/NIR+RED). Possible values of the dimensionless index are between –1 und +1. Values between 0.3-0.8 stand for a considerable photosynthetic activity in dense vegetation with a high biomass production. Soils with little vegetation cover tend to have small values. Here NDVI was used to represent biomass production in the different forest types along the

altitu-dinal gradient (WHITTAKER & NIERING 1975). FEELY et al. (2005), HE et al. (2009) and O L-DELAND et al. (2010) employed the NDVI for analyzing the floristic composition and struc-ture of forests.

Here NDVI was calculated on the May 3, 2003 Landsat 7 ETM+ scene downloaded from the ftp-server of the University of Maryland (ftp://ftp.glcf.umd.edu/glcf/Landsat/; Path: 008, Row: 047). On May 31, 2003 the Scan Line Corrector (SLC) of the Landsat 7 ETM+ failed so that all images taken after this date contain gapped data that increase towards the edge of a scene (SLC-off) (http://landsat.usgs.gov/Landsat_7_ETM_SLC_off_data_products.php). A functioning SLC would compensate for the forward motion of the satellite. As field work took place in February of 2011 a second SLC-off image with missing scan lines was downloaded from March 6, 2011. NDVI-values of the 2003 and the 2011 scene were compared for each plot. Six plots presented no data values in the 2011 scene as they coincided with missing data lines. The NDVI values of the plots on the two scenes were highly correlated (Pearson, r=0.727, p<0.001, N=137) and as tree composition was not considered to change rapidly the 2003 scene was used for further analysis. After downloading the image was registered to UTM coordinates (Zone 19N, WGS84) and referenced to a digital mosaic of the topographic maps of Armando Bermúdez National Park with a first order polynomial equation and nearest neighbor resampling technique. RMS error was smaller than 1 pixel.

Elevation is the primary source for deriving secondary environmental variables such as slope, aspect, hillshade and solar radiation, most widely used to determine differences in vegetation composition and prediction in mountainous areas, amongst other variables (W HITTAK-ER 1967; CAYUELA et al. 2006; SHERMAN et al. 2005). Elevation data were digitized as 100 m contour lines from topographic maps (1:50,000) at the Institute of Cartography, GIS and Re-mote Sensing at the University of Göttingen. The contour lines were interpolated to a 30 m spatial resolution digital elevation model (DEM) by the ArcInfo topogrid command integrat-ing mountain peaks and the hydrological network, digitized on topographic maps (H UTCHIN-SON 1988). Aspect, slope (in degrees) and hillshade were derived from the DEM. Slope is related to the hydrologic conditions, potential soil moisture and soil develop-ment (MOORE et al. 1991).

The hillshade was only used for reference of the relief, but not for analysis. All of the datasets were projected to UTM coordinates (Zone 19N, WGS84). To account for the different preci-pitation regime along the north-west and south-east facing slopes of the Cordillera Central one binary variable was calculated from aspect (north-east: 315-135° and south-west 135-315°). Furthermore aspect was transferred to two non-correlated linear variables representing

north-/southness (cos (radian)) and east-/westness (sin (radian)). Aspect is related to the radia-tion balance and potential evapotranspiraradia-tion of an area (MOORE et al. 1991).

As soil moisture is one of the most important determinants of vegetation composition (K O-PECKÝ & ČÍŽKOVÁ 2010; DIEKMANN 2003), the SAGA wetness index (SWI) was selected as a proxy. SWI is similar to the topographic wetness index (TWI), but works with a modified catchment area so that the grid cells in valley floors are assigned a more realistic, higher po-tential wetness (BÖHNER et al. 2002).

Another important factor for biological processes is incoming solar radiation. It was calcu-lated here from the DEM as the mean for the 12 months of 2003 (FU &RICH 2002).

The climatic regime was assessed by Worldclim data (HIJMANS et al. 2005) as high resolution meteorological data in continuous grids were not available in the Dominican Republic. The following climatic variables were downloaded from the Worldclim internet page (http://www.worldclim.org): annual mean temperature, mean diurnal temperature range (mean of monthly (max temp – min temp), temperature seasonality (standard deviation *100), annual precipitation, precipitation of wettest and driest month. As the three temperature variables were highly correlated (Pearson, all r > 0.8, p<0.001) and the three precipitation parameters also (Pearson, all r > 0.9, p<0.001), just annual mean temperature and annual precipitation were selected for further processing. To understand the effect of the trade wind inversion in the Worldclim precipitation model, it was plotted against the DEM (see Discussion of re-sults).

All datasets were masked with the DEM to the extent of the study area. Analyses were con-ducted at two spatial resolutions of 60 m x 60 m and 30 m x 30 m. As the results were similar the results for the 30 m x 30 m spatial resolution is presented here.

Processing was done with ERDAS Imagine 9.2 (ERDAS Inc., Atlanta, USA), SAGA-GIS 2.0 (SAGA User Group Association, Hamburg, Germany) and ArcGIS 9.3 (ESRI Inc., Redlands, USA).